Cost sensitive learning

Predicting underwriting risk has become a major challenge due to the imbalanced datasets in the field. A realworld imbalanced dataset is used in this work with 12 variables in 30144 cases, where most of the cases were classified as "accepting the insurance request", while a small percentage classified as "refusing insurance". This work developed 55 machine learning (ML) models to predict whether or not to renew policies. The models were developed using the original dataset and four data-level approaches resampling techniques: random oversampling, SMOTE, random undersampling, and hybrid methods with 11 ML algorithms to address the issue of imbalanced data (11 ML× (4 resampling techniques + unbalanced datasets) = 55 ML models). Seven classifier efficiency measures were used to evaluate these 55 models that were developed using 11 ML algorithms: logistic regression (LR), random forest (RF), artificial neural network (ANN), multilayer perceptron (MLP), support vector machine (SVM), naive Bayes (NB), decision tree (DT), XGBoost, k-nearest neighbors (KNN), stochastic gradient boosting (SGB), and AdaBoost. The seven classifier efficiency measures namely are accuracy, sensitivity, specificity, AUC, precision, F1-measure, and kappa. CRISP-DM methodology is utilisied to ensure that studies are conducted in a rigorous and systematic manner. Additionally, RapidMiner software was used to apply the algorithms and analyze the data, which highlighted the potential of ML to improve the accuracy of risk assessment in insurance underwriting. The results showed that all ML classifiers became more effective when using resampling strategies; where Hybrid resampling methods improved the performance of machine learning models on imbalanced data with an accuracy of 0.9967 and kappa statistics of 0.992 for the RF classifier.

Many practical data mining systems such as those for fraud detection and surveillance deal with building classifiers that are not autonomous but part of a larger interactive system with an expert in the loop. The goal of these systems is not just to maximize the performance of the classifier but to make the experts more efficient at performing their task, thus maximizing the overall Return on Investment of the system. This paper describes an interactive system for detecting payment errors in insurance claims with claim auditors in the loop. We describe an interactive claims prioritization component that uses an online cost-sensitive learning approach (more-like-this) to make the system efficient. Our interactive prioritization component is built on top of a batch classifier that has been trained to detect payment errors in health insurance claims and optimizes the interaction between the classifier and the domain experts who are consuming the results of this system. The goal is to make these auditors more efficient and effective as well as improving the classification performance of the system. The result is both a reduction in time it takes for the auditors to review and label claims as well as improving the precision of the system in finding payment errors. We show results obtained from applying this system at two major US health insurance companies indicating significant reduction in claim audit costs and potential savings of $20-$26 million/year making the insurance providers more efficient and lowering their operating costs. Our system reduces the money being wasted by providers and insurers dealing with incorrectly processed claims and makes the healthcare system more efficient.

Many practical data mining systems such as those for fraud detection and surveillance deal with building classifiers that are not autonomous but part of a larger interactive system with an expert in the loop. The goal of these systems is not just to maximize the performance of the classifier but to make the experts more efficient at performing their task, thus maximizing the overall Return on Investment of the system. This paper describes an interactive system for detecting payment errors in insurance claims with claim auditors in the loop. We describe an interactive claims prioritization component that uses an online cost-sensitive learning approach (more-like-this) to make the system efficient. Our interactive prioritization component is built on top of a batch classifier that has been trained to detect payment errors in health insurance claims and optimizes the interaction between the classifier and the domain experts who are consuming the results of this system. The goal is to make these auditors more efficient and effective as well as improving the classification performance of the system. The result is both a reduction in time it takes for the auditors to review and label claims as well as improving the precision of the system in finding payment errors. We show results obtained from applying this system at two major US health insurance companies indicating significant reduction in claim audit costs and potential savings of $20-$26 million/year making the insurance providers more efficient and lowering their operating costs. Our system reduces the money being wasted by providers and insurers dealing with incorrectly processed claims and makes the healthcare system more efficient.

It is often expensive to acquire data in real-world data mining applications. Most previous data mining and machine learning research, however, assumes that a fixed set of training examples is given. In this paper, we propose an online cost-sensitive framework that allows a learner to dynamically acquire examples as it learns, and to decide the ideal number of examples needed to minimize the total cost. We also propose a new strategy for Partial Example Acquisition (PAS), in which the learner can acquire examples with a subset of attribute values to reduce the data acquisition cost. Experiments on UCI datasets show that the new PAS strategy is an effective method in reducing the total cost for data acquisition.

Learning with different classification costs, cost-sensitive classification Cost-Sensitive Learning is a type of learning in data mining that takes the misclassification costs (and possibly other types of cost) into consideration. The goal of this type of learning is to minimize the total cost. The key difference between cost-sensitive learning and cost-insensitive learning is that cost-sensitive learning treats the different misclassifications differently. Costinsensitive learning does not take the misclassification costs into consideration. The goal of this type of learning is to pursue a high accuracy of classifying examples into a set of known classes. The class imbalanced datasets occurs in many real-world applications where the class distributions of data are highly imbalanced. Cost-sensitive learning is a common approach to solve this problem. Classification is the most important task in inductive learning and machine learning. A classifier can be trained from a set of training examples with class labels, and can be used to predict the class labels of new examples. The class label is usually discrete and finite. Many effective classification algorithms have been developed, such as naïve Bayes, decision trees, neural networks, and so on. However, most original classification algorithms pursue to minimize the error rate: the percentage of the incorrect prediction of class labels. They ignore the difference between types of misclassification errors. In particular, they implicitly assume that all misclassification errors cost equally. In many real-world applications, this assumption is not true. The differences between different misclassification errors can be quite large. For example, in medical diagnosis of a certain cancer, if the cancer is regarded as the positive class, and non-cancer (healthy) as negative, then missing a cancer (the patient is actually positive but is classified as negative; thus it is also called "false negative") is much more serious (thus expensive) than the false-positive error. The patient could lose his/her life because of the delay in the correct diagnosis and treatment. Similarly, if carrying a bomb is positive, then it is much more expensive to miss a terrorist who carries a bomb to a flight than searching an innocent person. Cost-sensitive learning takes costs, such as the misclassification cost, into consideration. It is one of the most active and important research areas in machine learning, and it plays an important

AI has the potential to greatly transform the
field of healthcare management through improving the
accuracy of diagnoses, effectiveness of treatments, and
overall experience of patients. This paper aims at
reviewing the utilization of four algorithms in the
healthcare sector namely; Decision Trees, Support
Vector Machines, Random Forests, and Neural
Networks. These algorithms are assessed on the basis of
performance indicators like accuracy, precision, recall
and F1 score using the patient record dataset of size
20,000 in the present work. Performance seems to show
that Neural Networks have the highest accuracy of 93
over the other algorithms. 5%, precision of 92. 0%, recall
of 93. 0 %, Precisions 92 %, Recal 93 % and F1 score of
92. 5%. This is followed by Random Forests with a 91.
3%, precision of 90. 0%, recall of 90. , accuracy of 5%,
and F1 score of 90. 2%. The research findings show that
SVM yields to 89 percent accuracy. 0%, precision of 87.
5%, recall of 88. 0%, Precision of 0.88 and F1 score of
0.874. for Decision Trees falls 85 % and for Naïve Bayes
it falls up to 73% whereas for Random Forest the
accuracy comes near to about 7%. 2%, precision of 82.
5%, recall of 84. The accuracy for the producers is 0%,
recall is 0%, and F1 score is 83. 2%. The comparative
analysis with the related works explores the pace and
progress of AI in various areas of the healthcare sector’s
concerns. Among the identified findings they include
proactive utilization of Artificial Intelligence in
enhancing the delivery of health care, use of Artificial
Intelligence in achieving patients’ better-quality health
care, and apply of Artificial Intelligence in the decision
making process. The future area is focused on the
improvement of the algorithm of AI, the ethical and legal
questions, and the usage of AI in the solving of the
worldwide health issues.

The past decade has seen a significant interest on the problem of inducing decision trees that take account of costs of misclassification and costs of acquiring the features used for decision making. This survey identifies over 50 algorithms including approaches that are direct adaptations of accuracy-based methods, use genetic algorithms, use anytime methods and utilize boosting and bagging. The survey brings together these different studies and novel approaches to cost-sensitive decision tree learning, provides a useful taxonomy, a historical timeline of how the field has developed and should provide a useful reference point for future research in this field.

Traditionally, software quality models are built by assuming a uniform misclassification cost. In other words, cost implications of misclassifying a fault prone module as fault free are assumed to be the same as the cost implications of misclassifying a fault free module as fault prone. In reality, these two types of misclassification costs are rarely equal. They are project-specific, reflecting the characteristics of the domain in which the program operates. In this paper, through the analysis of projects from a public repository, we analyze the benefits of techniques which incorporate misclassification costs in the development of software quality models. We find that cost-sensitive learning does not provide operational points which outperform cost-insensitive classifiers. However, an advantage of cost-sensitive modeling is the explicit choice of the operational threshold appropriate for the cost differential.

The usage of data is increasing day by day. There is a huge amount of data storage is required for handling the millions of twits, shares in social networks (twitter, facebook, WhatsApp, and youtube) per second. Databases are playing a vital role in data warehousing and mining. The process of storing the data in large repository place is known as Data Warehousing. Nowadays, Search Engines are struggling to follow the Search Engine Optimization techniques. So there is a pressure for the data analyst to fetch the data from the data warehouse efficiently. The task of classification with imbalanced datasets have attracted quite an interest from researchers in the recent years. Accordingly, various classification techniques are used to handle the newly arrived large amount of data. So many applications have been designed to address this problem from the different perspective such as data pre-processing, algorithm modification and sensitive learning. The problem of constructing fast and a...

Artificial intelligence (AI) is increasingly integrated into healthcare systems, promising advancements in diagnostics, treatment optimization, and patient care. However, patient apprehensions regarding AI adoption persist, influenced by concerns over privacy, data security, ethical implications, and the potential for AI to replace human interaction in healthcare delivery. This review explores patient perspectives on AI in healthcare, highlighting key apprehensions and factors influencing patient acceptance or reluctance. Understanding these concerns is crucial for developing ethical guidelines, ensuring transparent AI deployment, and fostering patient-centered AI implementations that prioritize trust, safety, and equitable access to healthcare innovations.

In the states, penetration of mobile platforms at an equal speed is completely changing things around for people and enterprises making life easy and accessible. This move, however has new baggage of cyber threats associated with it that made mobile devices the softest targets for an attack. The present study sheds a light on the strategies and inventions demanded in mobile platform cybersecurity evolution after analyzing contemporary literature to uncover latent loopholes inside U.S. Mobile platforms, namely application security; data encryption, and user authentication unit. More report on Artificial Intelligence (AI), Machine Learning(ML) and Blockchain as the technology can be used to have better security in mobile. Exhaustive evaluation of case studies and trends emerging in the market is presented to understand how these technologies have been proven beneficial across various use cases. This highlights again how crucial it is to take a multi-layered approach when tackling mobile security, combining innovative technologies with strong policies and effective user education. The paper concludes by providing specific strategic recommendations for actors in the mobile ecosystem and highlights how continuous innovation and collaboration are key to stay one step ahead of cybercriminals as they adapt.

Cyber threats targeting these systems are increasingly being refined as the use of online banking platforms continue to grow. To address these concerns, this paper provides an insightful idea on how integrating AI and ML algorithms can help to boost the cybersecurity in online banking platforms of US. Our advanced security system is comprised of intelligent and adaptive AIpowered anomaly detection, dynamic ML based predictive analytics, automated response that provides defense automate intrusion prevention systems in the cloud which brings scale with ease managed services providing better fraud analytical predictions combined with a real-time threat Intelligence suite. To address this gap, we conduct an extensive survey of the available literature to highlight important AI and ML methodologies that can be used in online banking including supervised algorithms which could be applied for transaction monitoring, unsupervised learning strategies can potentially provide better results but at higher computational costs concerning anomaly detection techniques as well as reinforcement learning based approaches that might assist on dynamic security measures. We also offer a case study that shows the deployment of these technologies in an online banking simulation and illustrates their efficacy for detecting and preventing security violations.

Machine learning model validation is a cornerstone and a requisite before a machine learning model can be generalized or relied upon. This paper analyses the validation strategy challenges and solutions to quantify cross validation methodologies, to employ appropriate data splitting techniques, and to employ proper validation approaches for various data types. These aspects must be understood in order to develop robust and reliable machine learning models.

Predicting hospital readmission is crucial for improving patient care and optimizing healthcare resource allocation. Traditional methods often overlook the imbalanced costs associated with different types of prediction errors. This study proposes a cost-sensitive deep learning approach tailored for hospital readmission prediction. By integrating cost-sensitive learning techniques into deep neural networks, the model prioritizes minimizing costly prediction errors, such as false negatives. Experimental results on a large-scale healthcare dataset demonstrate the effectiveness of the proposed approach in achieving higher predictive accuracy while mitigating readmission risks. The findings highlight the potential of cost-sensitive deep learning to enhance healthcare outcomes and resource utilization.

Brain-computer interface (BCI) technology holds promise for revolutionizing interaction between humans and machines, enabling direct communication pathways based on neural signals. However, optimizing BCI applications requires overcoming challenges such as signal noise, latency, and system integration complexities. This paper proposes leveraging Internet of Things (IoT) technologies to enhance BCI applications through real-time data processing, adaptive feedback mechanisms, and seamless connectivity. By integrating IoT with BCI systems, this approach aims to improve usability, reliability, and performance, thereby advancing the capabilities and accessibility of brain-controlled devices in various domains including healthcare, gaming, and assistive technology.

In the realm of pervasive computing, where interconnected devices seamlessly integrate into everyday environments, the role of middleware becomes pivotal in managing and leveraging contextual information. This study proposes a reconfigurable context-sensitive middleware framework designed to dynamically adapt to changing environmental contexts and user interactions. By incorporating context awareness into its design, the middleware enhances the interoperability and intelligence of pervasive computing systems, facilitating efficient resource management and personalized services delivery. Key functionalities include context acquisition, interpretation, and adaptation, enabling the middleware to optimize device interactions based on situational awareness. Through a comprehensive evaluation, including performance metrics and usability assessments, the framework demonstrates its efficacy in enhancing system flexibility, responsiveness, and user experience in diverse pervasive computing scenarios.

In the process of bankruptcy prediction models, a class imbalanced problem has occurred which limits the performance of the models. Most prior research addressed the problem by applying resampling methods such as the synthetic minority oversampling technique (SMOTE). However, resampling methods lead to other issues, e.g., increasing noisy data and training time during the process. To improve the bankruptcy prediction model, we propose cost-sensitive extreme gradient boosting (CS-XGB) to address the class imbalanced problem without requiring any resampling method. The proposed method’s effectiveness is evaluated on six real-world datasets, i.e., the LendingClub, and five Polish companies’ bankruptcy. This research compares the performance of CS-XGB with other ensemble methods, including SMOTE-XGB which applies SMOTE to the training set before the learning process. The experimental results show that i) based on LendingClub, the CS-XGB improves the performance of XGBoost and SMOTE-XGB ...

Existing supervised learning techniques can support product recommendations but are ineffective in scenarios characterized by single-class learning; i.e., training samples consisted of some positive examples and a much greater number of unlabeled examples. To address the limitations inherent in existing single-class learning techniques, we develop COst-sensitive Learning-based Positive Example Learning (COLPEL), which constructs an automated classifier from a singleclass training sample. Our method employs cost-proportionate rejection sampling to derive, from unlabeled examples, a subset likely to feature negative examples, according to the respective misclassification costs. COLPEL follows a committee machine strategy, thereby constructing a set of automated classifiers used together to reduce probable biases common to a single classifier. We use customers' book ratings from the Amazon.com Web site to evaluate COLPEL, with PNB and PEBL as benchmarks. Our results show that COLPEL outperforms both PNB and PEBL, as measured by its accuracy, positive F1 score, and negative F1 score.

Parkinson's disease is a neurodegenerative condition that affects billions of persons worldwide. This abstract aims to shed light on the causes and consequences of this debilitating condition. The primary cause of Parkinson's disease is the progressive degeneration of dopaminergic neurons in the substantia nigra region of brain. This neuronal loss results in a depletion of dopamine, a crucial neurotransmitter responsible for regulating movement and coordination. Therefore, individuals with Parkinson's disease have symptoms like tremors, rigidity, bradykinesia, and postural instability. These signs profoundly impact the quality of life, causing difficulties with daily activities and reducing independence. In addition to motor symptoms, non-motor symptoms such as depression, cognitive impairment, and autonomic dysfunction often accompany the disease, further complicating the clinical picture. Research into the causes and consequences of Parkinson's disease is ongoing, with a focus on using efficient medications and refining the quality of life for those affected by this condition. Now by Using machine learning algorithms, we can predict whether a person has a specific disease based on input values like gender and age. These algorithms analyze patterns and relationships in data to get predictions about an individual's health status. This technology can assist in early disease detection and improve healthcare outcomes..

Classification of data has become an important research area. The process of classifying documents into predefined categories Unbalanced data set, a problem often found in real world application, can cause seriously negative effect on classification performance of machine learning algorithms. There have been many attempts at dealing with classification of unbalanced data sets. In this paper we present a brief review of existing solutions to the class-imbalance problem proposed both at the data and algorithmic levels. Even though a common practice to handle the problem of imbalanced data is to rebalance them artificially by oversampling and/or under-sampling, some researchers proved that modified support vector machine, rough set based minority class oriented rule learning methods, cost sensitive classifier perform good on imbalanced data set. We observed that current research in imbalance data problem is moving to hybrid algorithms.

Boosting is one of the most important recent developments in classification methodology. It can significantly improve the prediction performance of any single classification algorithm and has been successfully applied to many different fields including problems in chemometrics. Boosting works by sequentially applying a classification algorithm to reweighted versions of the training data, and then taking a weighted majority vote of the sequence of classifiers thus produced. In this paper, we proposed a generalized boosting algorithm via Bayes optimal decision rule. Using Bayes optimal decision rule, we adjust the weights of the sequence of classifiers in the voting process of boosting algorithm. The two types of errors are introduced into the generalized boosting and make the voting process more sensible. Meanwhile, the weights of the training samples are also correspondingly adjusted according to some criterion. The generalized boosting is applied to the binary classification for chemical data. Experimental results show that it can improve the predict accuracy compared with AdaBoost algorithm especially when the difference between the two types of errors for classification is large.

Health datasets typically comprise of data that are heavily skewed towards the healthy class, thus resulting in classifiers being biased towards this majority class. Due to this imbalance of data, traditional performance metrics, such as accuracy, are not appropriate for evaluating the performance of classifiers with the minority class (disease-affected/unhealthy individuals). In addition, classifiers are trained under the assumption that the costs or benefits associated with different decision outcomes are equal. However, this is usually not the case with health data since it is more important to identify disease affected/unhealthy persons rather than healthy individuals. In this paper we address these problems by examining benefits/costs when evaluating the performance of classifiers. Furthermore, we focus on multiclass classification where the outcome can be one of three or more options. We propose modifications to the Naive Bayes and Logistic Regression algorithms to incorporate...

The use of classification methods in real-world problems has costs that are usually neglected in the early algorithms which cause inefficiencies in practice. One of these costs, which is significant in many cases, is the cost of obtaining feature values for each instance, named Test-Cost. The Ensemble of classifiers as a common and practical classification method, is also considered and used in this perspective. Each classifier needs a number of features to classify the sample; if instead of using all classifiers, the best arrange of classifiers with the aim of minimizing the needed features is found, an effective solution for lowering the test-cost is obtained. In this paper, a method is proposed which uses reinforcement learning to construct such a Classifier Ensemble. The proposed method learns to find the best sequence of classifiers for each sample to minimize the test-cost. Two problems, an easy one and a hard one, are considered for testing the proposed method, in both of which yields very good results.

Diabetes mellitus is a hyperglycemia-like chronic condition that is a troublesome disease. It is estimated that, according to the growing morbidity, by 2040, the world will cross 642 million diabetic patients. This means that each one of the ten adults will be diabetes-affected. Diabetes can also lead to other illnesses such as heart attacks, kidney damage, and even blindness. The prediction of diabetes in advance motivates us to develop a machine learning-based model. A dataset was obtained from the online repository for this work. The obtained dataset was imbalanced. An imbalanced dataset presents a challenge that is needed to be balanced for prediction using multiple machine learning like Tomek and SMOTE. These techniques remove necessary outliers that are incomplete in the provided dataset. These outliers are also managed using the IQR method. Additionally, this research employed a two-stage model selection methodology. In the first stage, logistic regression, Support Vector Mac...

Cost-sensitive learning which deals with classification problems that have non-uniform costs has attracted great attention from the machine learning and data mining communities in recent years. In this study, a rescaling based meta-learning scheme is applied to cost-insensitive MEPAR-miner and DIFACONN-miner algorithms which were previously developed by the authors in order to make the algorithms cost-sensitive. Rescaling is realized in two ways by means of oversampling and undersampling by resampling the training instances in proportion to their costs. The proposed algorithms can extract rules for both binary and n-ary classification problems and also handle data sets that have missing values. An extensive computational study is performed on different types of classification benchmark problems with the aim of testing the performances of the algorithms. Comparisons with traditional cost-sensitive meta-learning algorithms and cost-insensitive MEPAR-miner and DIFACONN-miner algorithms show that the proposed cost-sensitive algorithms are competitive meta-learning algorithms and able to produce accurate and effective classification rules with low misclassification costs.

Classical construction of ecological models follows one of the following two approaches: (1) either the measured data are analyzed with a statistical approach and a black-box statistical model is constructed, or (2) the model is deduced from basic (physical, chemical, etc.) principles and the measured data are used to calibrate and validate the model. A combination of the both ways is used very rarely, although it could combine the benefits of both. Machine learning tools (ML), developed within the area of artificial intelligence (AI), are able to analyze a data base (perform data mining), using also some domain (expert) knowledge. As a result, ML tools can automatically, or with limited expert help, induce compact and easy to understand models from measured data. In this paper a description of the ML tool GoldHorn that induces differential or difference (algebraic) equations from measured time series data is given. Its successful application to the prediction of algae growth in the...

The use of classification as a data mining approach for performance prediction has been studied by many eminent researchers. The objective of this study is to determine the best classification models for predicting At Risk status of students in their first semester of an undergraduate degree program. A comprehensive evaluation requires that multiple models with different algorithms were analyzed using key performance measures. Principal component analysis and feature selection by weights using information gain ratio, Gini index, correlation and PCA is used to determine the relevant predictors of the datasets used. This study also addresses gaps in the current available literature on performance prediction, such as data imbalance and the use of Ensemble models. Sampling and weighting techniques were included using Rapid Miner operators for SMOTE, stratified, bootstrap sampling and weighting. Ensemble models using bagging, boosting and the vote operator in addition to Gradient Boosted trees and Random Forest were compared to the individual classifiers to measure model efficiency. The best models were then used to ascertain how early at-risk prediction can be employed using data on student performance in the course assessments. The results show that Ensemble models and the use of sampling and weighting clearly improves model performance. The early risk prediction as expected is most accurate with all the coursework and final grades in a semester. Interestingly the variance in the performance measure values are not very significant for some of the models and it can be concluded that early risk prediction can happen earlier in the semester when intervention and associated benefits of improving student performance is more probable.

Loan fraud is a critical factor in the insolvency of financial institutions, so companies make an effort to reduce the loss from fraud by building a model for proactive fraud prediction. However, there are still two critical problems to be resolved for the fraud detection: (1) the lack of cost sensitivity between type I error and type II error in most prediction models, and (2) highly skewed distribution of class in the dataset used for fraud detection because of sparse fraud-related data. The objective of this paper is to examine whether classification cost is affected both by the cost-sensitive approach and by skewed distribution of class. To that end, we compare the classification cost incurred by a traditional cost-insensitive classification approach and two cost-sensitive classification approaches, Cost-Sensitive Classifier (CSC) and Meta-Cost. Experiments were conducted with a credit loan dataset from a major financial institution in Korea, while varying the distribution of class in the dataset and the number of input variables. The experiments showed that the lowest classification cost was incurred when the MetaCost approach was used and when non-fraud data and fraud data were balanced. In addition, the dataset that includes all delinquency variables was shown to be most effective on reducing the classification cost.

Conformal prediction (CP) is a wrapper around traditional machine learning models, giving coverage guarantees under the sole assumption of exchangeability; in classification problems, a CP guarantees that the error rate is at most a chosen significance level ε, irrespective of whether the underlying model is misspecified. However, the prohibitive computational costs of full CP led researchers to design scalable alternatives, which alas do not attain the same guarantees or statistical power of full CP. In this paper, we use influence functions to efficiently approximate full CP. We prove that our method is a consistent approximation of full CP, and empirically show that the approximation error becomes smaller as the training set increases; e.g., for 1, 000 training points the two methods output p-values that are < 0.001 apart: a negligible error for any practical application. Our methods enable scaling full CP to large real-world datasets. We compare our full CP approximation (ACP) to mainstream CP alternatives, and observe that our method is computationally competitive whilst enjoying the statistical predictive power of full CP. Method Prediction set ACP bird, cat, deer, frog SCP bird, deer, frog RAPS bird, cat, deer, dog, frog CV+ bird, cat, deer, dog, frog Method Prediction set ACP auto, cat, frog, horse, truck SCP auto, deer, frog, truck RAPS plane, auto, bird, deer, frog, ship, truck CV+ plane, auto, deer, frog, horse, truck Method Prediction set ACP cat, deer, frog, horse SCP cat, deer, dog, frog, horse, truck RAPS cat, deer, dog, frog, horse, truck CV+ cat, deer, dog, frog, horse

The multiclass imbalanced data problems in data mining were interesting cases to study currently. The problems had an influence on the classification process in machine learning processes. Some cases showed that minority class in the dataset had an important information value compared to the majority class. When minority class was misclassification, it would affect the accuracy value and classifier performance. In this research, cost sensitive decision tree C5.0 was used to solve multiclass imbalanced data problems. The first stage, making the decision tree model uses the C5.0 algorithm then the cost sensitive learning uses the metacost method to obtain the minimum cost model. The results of testing the C5.0 algorithm had better performance than C4.5 and ID3 algorithms. The percentage of algorithm performance from C5.0, C4.5 and ID3 were 40.91%, 40, 24% and 19.23%

Discretization is defined as the process that divides continuous numeric values into intervals of discrete categorical values. In this article, the concept of cost-based discretization as a pre-processing step to the induction of a classifier is introduced in order to obtain an optimal multi-interval splitting for each numeric attribute. A transparent description of the method and the steps involved in cost-based discretization are given. The aim of this paper is to present this method and to assess the potential benefits of such an approach. Furthermore, its performance against two other well-known methods, i.e. entropy-and pure error-based discretization is examined. To this end, experiments on 14 data sets, taken from the UCI Repository on Machine Learning were carried out. In order to compare the different methods, the area under the Receiver Operating Characteristic (ROC) graph was used and tested on its level of significance. For most data sets the results show that cost-based discretization achieves satisfactory results when compared to entropy-and error-based discretization. Given its importance, many researchers have already contributed to the issue of discretization in the past. However, to the best of our knowledge, no efforts have been made yet to include the concept of misclassification costs to find an optimal multi-split for discretization purposes, prior to induction of the decision tree. For this reason, this new concept is introduced and explored in this article by means of operations research techniques.

Heart disease, one of the major causes of mortality worldwide, can be mitigated by early heart disease diagnosis. A clinical decision support system (CDSS) can be used to diagnose the subjects' heart disease status earlier. This study proposes an effective heart disease prediction model (HDPM) for a CDSS which consists of Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to detect and eliminate the outliers, a hybrid Synthetic Minority Over-sampling Technique-Edited Nearest Neighbor (SMOTE-ENN) to balance the training data distribution and XGBoost to predict heart disease. Two publicly available datasets (Statlog and Cleveland) were used to build the model and compare the results with those of other models (naive bayes (NB), logistic regression (LR), multilayer perceptron (MLP), support vector machine (SVM), decision tree (DT), and random forest (RF)) and of previous study results. The results revealed that the proposed model outperformed other models and previous study results by achieving accuracies of 95.90% and 98.40% for Statlog and Cleveland datasets, respectively. In addition, we designed and developed the prototype of the Heart Disease CDSS (HDCDSS) to help doctors/clinicians diagnose the patients'/subjects' heart disease status based on their current condition. Therefore, early treatment could be conducted to prevent the deaths caused by late heart disease diagnosis. INDEX TERMS Heart disease, disease prediction model, clinical decision support system, outlier data, imbalanced data, machine learning.

The global need for effective disease diagnosis remains substantial, given the complexities of various disease mechanisms and diverse patient symptoms. To tackle these challenges, researchers, physicians, and patients are turning to machine learning (ML), an artificial intelligence (AI) discipline, to develop solutions. By leveraging sophisticated ML and AI methods, healthcare stakeholders gain enhanced diagnostic and treatment capabilities. However, there is a scarcity of research focused on ML algorithms for enhancing the accuracy and computational efficiency. This research investigates the capacity of machine learning algorithms to improve the transmission of heart rate data in time series healthcare metrics, concentrating particularly on optimizing accuracy and efficiency. By exploring various ML algorithms used in healthcare applications, the review presents the latest trends and approaches in ML-based disease diagnosis (MLBDD). The factors under consideration include the algorithm utilized, the types of diseases targeted, the data types employed, the applications, and the evaluation metrics. This review aims to shed light on the prospects of ML in healthcare, particularly in disease diagnosis. By analyzing the current literature, the study provides insights into state-of-the-art methodologies and their performance metrics.

We design a new adaptive learning algorithm for misclassification cost problems that attempt to reduce the cost of misclassified instances derived from the consequences of various errors. Our algorithm (adaptive cost sensitive learning-AdaCSL) adaptively adjusts the loss function such that the classifier bridges the difference between the class distributions between subgroups of samples in the training and test data sets with similar predicted probabilities (i.e., local trainingtest class distribution mismatch). We provide some theoretical performance guarantees on the proposed algorithm and present empirical evidence that a deep neural network used with the proposed AdaCSL algorithm yields better cost results on several binary classification data sets that have class-imbalanced and classbalanced distributions compared to other alternative approaches.

Cost-sensitive learning based on Bregman divergences Santos-Rodríguez et al. 1/ 15 Outline 1 Introduction 2 Posterior probability estimation 3 Designing Bregman Divergences 4 Towards a maximum margin classifier 5 Conclusions Cost-sensitive learning based on Bregman divergences Santos-Rodríguez et al. 2/ 15 Decision costs → c(d, d) d: decision (based on sample x) d: true hypothesis. Goal minimizing the mean risk E {c(d, d)} Cost-sensitive learning based on Bregman divergences Santos-Rodríguez et al. 3/ 15 Cost-sensitive learning based on Bregman divergences Santos-Rodríguez et al. 4/ 15 Cost-sensitive learning based on Bregman divergences Santos-Rodríguez et al. 5/ 15 Cost-sensitive learning based on Bregman divergences Santos-Rodríguez et al. 6/ 15 Cost-sensitive learning based on Bregman divergences Santos-Rodríguez et al. 9/ 15 Cost-sensitive learning based on Bregman divergences Santos-Rodríguez et al.

A lot of approaches, each following a different strategy, have been proposed in the literature to provide AdaBoost with cost-sensitive properties. In the first part of this series of two papers, we have presented these algorithms in a homogeneous notational framework, proposed a clustering scheme for them and performed a thorough theoretical analysis of those approaches with a fully theoretical foundation. The present paper, in order to complete our analysis, is focused on the empirical study of all the algorithms previously presented over a wide range of heterogeneous classification problems. The results of our experiments, confirming the theoretical conclusions, seem to reveal that the simplest approach, just based on cost-sensitive weight initialization, is the one showing the best and soundest results, despite having been recurrently overlooked in the literature.

Classification is one important area in machine learning that labels the class of an instance via a classifier from known-class historical data. One of the popular classifiers is k-NN, which stands for “k-nearest neighbor” and requires a global parameter k to proceed. This global parameter may not be suitable for all instances. Naturally, each instance may situate on different regions of clusters such as an interior instance placed inside a cluster, a border instance placed on the outskirts, an outer instance placed faraway from any cluster, which requires a different number of neighbors. To automatically assign a different number of neighbors to each instance, the concept of scoring from the anomaly detection research is desired. The Mass-ratio-variance Outlier Factor, MOF, is selected as the scoring scheme for the number of neighbors of each instance. MOF gives the highest score to an instance placed very far from any cluster and the lowest score to an instance surrounded by other...

The paper considers a conformal prediction method for bounded regression task. A predictor was based on the Defensive Forecast algorithm and has been applied for a medical prognostic problem. These empirical results are compared and discussed.

The work describes an application of a recently developed machine-learning technique called Mondrian predictors to risk assessment of ovarian and breast cancers. The analysis is based on mass spectrometry profiling of human serum samples that were collected in the United Kingdom Collaborative Trial of Ovarian Cancer Screening. The work describes the technique and presents the results of classification (diagnosis) and the corresponding measures of confidence of the diagnostics. The main advantage of this approach is a proven validity of prediction. The work also describes an approach to improve early diagnosis of ovarian and breast cancers since the data in the United Kingdom Collaborative Trial of Ovarian Cancer Screening were collected over a period of 7 years and do allow to make observations of changes in human serum over that period of time. Significance of improvement is confirmed statistically (for up to 11 months for ovarian cancer and 9 months for breast cancer). In addition, the methodology allowed us to pinpoint

Semi-supervised learning is one of the significant field in machine learning or data mining. It deals with datasets that have many unlabeled and a few labeled samples. In this study we aim to predict students' success in educational institutes by use of semi-supervised classification methods in the wellknown machine learning tool WEKA. The methods are explained in detail and for every method, implementations are done on a special dataset called "Students' performance". The effectiveness of the methods are tried to be increased by using attrbibute selection functions in data selection and transformation processes. The performance of the algorithms are stated by accuracy results presented in tables and figures.

In the last years, it has become critical for financial institutions to develop efficient solutions that leverage all available heterogeneous customer data. While financial crime is growing in scale and complexity, financial analysts require a new set of tools to protect and shield the financial system against illegal activities. We propose a multi-view semi-supervised risk classification approach that makes use of text and financial transactions data for Customer Due Diligence (CDD) reporting. We conducted a case study which illustrates that by using multiple learning sources (views) at the same time, higher performance is achieved compared to models learned on each source individually. Furthermore, our system allows to explain why and how the prediction was done thus providing new insight to domain experts.

This paper presents fraud detection problem as one of the most common problems in secure banking research field, due to its importance in reducing the losses of banks and e-transactions companies. Our work will include: applying the common classification algorithms such as logistic regression (LR), random forest (RF), alongside with modern classifiers with state-of-the-art results as XGBoost (XG) and CatBoost (CB), testing the effect of the unbalanced data through comparing their results with and without balancing, then focusing on the savings measure to test the effect of cost-sensitive wrapping of Bayes minimum risk (BMR), we will concentrate on using F1-score, AUC and Savings measures after using the traditional measures duo to their suitability to our problem. The results show that CB has the best savings (0.7158) alone, (0.971) when using SMOTE and (0.9762) with SMOTE and BMR, while XG has the best savings (0.757) when using BMR without SMOTE.

Heart disease, one of the major causes of mortality worldwide, can be mitigated by early heart disease diagnosis. A clinical decision support system (CDSS) can be used to diagnose the subjects' heart disease status earlier. This study proposes an effective heart disease prediction model (HDPM) for a CDSS which consists of Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to detect and eliminate the outliers, a hybrid Synthetic Minority Over-sampling Technique-Edited Nearest Neighbor (SMOTE-ENN) to balance the training data distribution and XGBoost to predict heart disease. Two publicly available datasets (Statlog and Cleveland) were used to build the model and compare the results with those of other models (naive bayes (NB), logistic regression (LR), multilayer perceptron (MLP), support vector machine (SVM), decision tree (DT), and random forest (RF)) and of previous study results. The results revealed that the proposed model outperformed other models and previous study results by achieving accuracies of 95.90% and 98.40% for Statlog and Cleveland datasets, respectively. In addition, we designed and developed the prototype of the Heart Disease CDSS (HDCDSS) to help doctors/clinicians diagnose the patients'/subjects' heart disease status based on their current condition. Therefore, early treatment could be conducted to prevent the deaths caused by late heart disease diagnosis. INDEX TERMS Heart disease, disease prediction model, clinical decision support system, outlier data, imbalanced data, machine learning.

Early diseases prediction plays an important role for improving healthcare quality and can help individuals avoid dangerous health situations before it is too late. This paper proposes a disease prediction model (DPM) to provide an early prediction for type 2 diabetes and hypertension based on individual's risk factors data. The proposed DPM consists of isolation forest (iForest) based outlier detection method to remove outlier data, synthetic minority oversampling technique tomek link (SMOTETomek) to balance data distribution, and ensemble approach to predict the diseases. Four datasets were utilized to build the model and extract the most significant risks factors. The results showed that the proposed DPM achieved highest accuracy when compared to other models and previous studies. We also developed a mobile application to provide the practical application of the proposed DPM. The developed mobile application gathers risk factor data and send it to a remote server, so that an individual's current condition can be diagnosed with the proposed DPM. The prediction result is then sent back to the mobile application; thus, immediate and appropriate action can be taken to reduce and prevent individual's risks once unexpected health situations occur (i.e., type 2 diabetes and/or hypertension) at early stages. INDEX TERMS Diabetes, disease prediction, ensemble learning, hypertension, imbalanced data, outlier data.

We provide a unifying perspective for two decades of work on cost-sensitive Boosting algorithms. When analyzing the literature 1997-2016, we find 15 distinct costsensitive variants of the original algorithm; each of these has its own motivation and claims to superiority-so who should we believe? In this work we critique the Boosting literature using four theoretical frameworks: Bayesian decision theory, the functional gradient descent view, margin theory, and probabilistic modelling. Our finding is that only three algorithms are fully supported-and the probabilistic model view suggests that all require their outputs to be calibrated for best performance. Experiments on 18 datasets across 21 degrees of imbalance support the hypothesis-showing that once calibrated, they perform equivalently, and outperform all others. Our final recommendation-based on simplicity, flexibility and

In cost-sensitive learning, misclassification costs can vary for different classes. This paper investigates an approach reducing a multi-class cost-sensitive learning to a standard classification task based on the data space expansion technique developed by Abe et al., which coincides with Elkan's reduction with respect to binary classification tasks. Using this proposed reduction approach, a cost-sensitive learning problem can be solved by considering a standard 0/1 loss classification problem on a new distribution determined by the cost matrix. We also propose a new weighting mechanism to solve the reduced standard classification problem, based on a theorem stating that the empirical loss on independently identically distributed samples from the new distribution is essentially the same as the loss on the expanded weighted training set. Experimental results on several synthetic and benchmark datasets show that our weighting approach is more effective than existing representative approaches for cost-sensitive learning.

This research explores Cost-Sensitive Learning (CSL) in the fraud detection domain to decrease the fraud class's incorrect predictions and increase its accuracy. Notably, we concentrate on shill bidding fraud that is challenging to detect because the behavior of shill and legitimate bidders are similar. We investigate CSL within the Semi-Supervised Classification (SSC) framework to address the scarcity of labeled fraud data. Our paper is the first attempt to integrate CSL with SSC for fraud detection. We adopt a meta-CSL approach to manage the costs of misclassification errors, while SSC algorithms are trained with imbalanced data. Using an actual shill bidding dataset, we assess the performance of several hybrid models of CSL and SSC and then compare their misclassification error and accuracy rates statistically. The most efficient CSL+SSC model was able to detect 99% of fraudsters and with the lowest total cost.

This paper describes modelling of time behaviour of phytoplankton and zooplankton in the Danish lake Glumsø with a recently developed approach to machine learning in numerical domains, called Q 2 learning. An essential part of this approach is qualitative constraints which were either handcrafted using knowledge from the Lotka-Volterra predator-prey model or induced directly from the collected data with a program called QUIN. The induced models were evaluated by a domain expert. We performed a comparison between numerical results of the Q 2 learning approach and standard machine learning algorithms. The results suggest that use of qualitative constraints leads to more accurate quantitative predictions.