Feature Projection Based Rule Classification

Rule based classification or rule induction (RI) in data mining is an approach that normally generates classifiers containing simple yet effective rules. Most RI algorithms suffer from few drawbacks mainly related to rule pruning and rules sharing training data instances. In response to the above two issues, a new dynamic rule induction (DRI) method is proposed that utilises two thresholds to minimise the items search space. Whenever a rule is generated, DRI algorithm ensures that all candidate items' frequencies are updated to reflect the deletion of the rule's training data instances. Therefore, the remaining candidate items waiting to be added to other rules have dynamic frequencies rather static. This enables DRI to generate not only rules with 100% accuracy but rules with high accuracy as well. Experimental tests using a number of UCI data sets have been conducted using a number of RI algorithms. The results clearly show competitive performance in regards to classification accuracy and classifier size of DRI when compared to other RI algorithms.

2010

Algorithms for feature selection in predictive data mining for classification problems attempt to select those features that are relevant, and are not redundant for the classification task. A relevant feature is defined as one which is highly correlated with the target function. One problem with the definition of feature relevance is that there is no universally accepted definition of what it means for a feature to be 'highly correlated with the target function or highly correlated with the other features'. A new feature selection algorithm which incorporates domain specific definitions of high, medium and low correlations is proposed in this paper. The proposed algorithm conducts a heuristic search for the most relevant features for the prediction task.

2005

Abstract—Classification Rule Mining (CRM) is a Data Mining technique for the extraction of hidden Classification Rules (CRs) from a given database, the objective being to build a classifier to classify “unseen” data. One recent approach to CRM is to use Association Rule Mining (ARM) techniques to identify the desired CRs, ie Classification Association Rule Mining (CARM).

IJCSMC, 2019

The interesting association rules is a special part of knowledge extraction from data. Apriori's support-and rule-based algorithms have provided an elegant solution to the problem of rule mining, but they produce too much rules, selecting some rules of no interest and ignoring rules[3] [5]. interesting. Other measures are needed to complete the support and the confidance. In this paper, we review the main measures proposed in the literature and we propose criteria to evaluate them. We then suggest a validation method that uses the tools of statistical learning theory, including VC-dimension. Given the large number of measurements and the multitude of candidate rules, the interest of these tools is to allow the construction of uniform non-asymptotic terminals for all the rules and all the measurements simultaneously.

1999

Abstract This paper presents a measure which estimates interestingness of a rule relative to its corresponding common sense rules. Mining interesting rules is one of the important data mining tasks. Interesting rules bring novel knowledge that helps decision makers for advantageous actions. Interestingness is a relative issue. It is relative with what is known about the domain. A measure which can estimate the interestingness of a rule relative to the known knowledge is thus required.

IEEE Access

Data mining has become a popular process in recent times. However, with the increase in data, traditional data mining methods are not sufficient to solve many problems. Therefore, advanced techniques are needed to provide better results without consuming more time during execution. Soft computing algorithms are used for mathematical optimization to achieve better results in less time. The primary purpose of this work is to propose a framework for rule mining that shall generalize the currently applied methods in rule mining. In this respect, this paper represents the R-miner using a soft computing algorithm. The Grouped-Artificial Bee Colony Optimization (G-ABC) was used to select the relevant attribute set and further verify the features. Mean-Variance optimization is used to find whether the selected rule is valid for further classification. Furthermore, a neural-based deep learning method is applied to validate the outcome. The investigation outcome indicates that the proposed algorithm provides more optimized results in terms of the number of rules generated, the time required for calculation, and obtaining supplementary information for rule mining. INDEX TERMS Rule mining, feature selection, particle swarm optimization, artificial bee colony optimization.

Journal of Convergence Information Technology, 2010

Association rule mining process can be visualized as a multi-objective problem rather than as a single objective one. Measures like support, confidence and other interestingness criteria which are used for evaluating a rule, can be thought of as different objectives of association rule mining problem. Support count is the number of records, which satisfies all the conditions that exist in the rule. This objective represents the accuracy of the rules extracted from the database. Confidence represents the proportion of records for which the prediction of the rule (or model in the case of a complete classification) is correct, and it is one of the most widely quoted measures of quality, especially in the context of complete classification. Interestingness measures how much interesting the rule is. Using these three measures as the objectives of rule mining problem, this article uses a Simulated Annealing algorithm to extract some useful and interesting rules from any type databases. The experimental results show that the algorithm may be suitable for large datasets.

2018

There are two main approaches for feature subset selection, i.e., wrapper and filter based. In wrapper based approach, which is a supervised method, the feature subset selection algorithm acts as a wrapper around an induction algorithm. The induction algorithm is actually a black-box for the feature subset selection algorithm and is mostly the classifier itself. The filter approach is an unsupervised method and attempts to assess the merits of features from the data while ignoring the performance of the induction algorithm. In this study, the effects of the feature subset selection approaches on the classification performance of rule-based learning algorithms, i.e., C4.5, RIPPER, PART, BFTree were investigated. These algorithms are fast in case of wrapper based approach. For various datasets, significant accuracy improvements were achieved with the wrapper based feature subset selection method. Other algorithms like Multilayer Perceptron (MLP) and Random Forests (RF) were also appli...

Lecture Notes in Computer Science, 2006

Inducing classification rules on domains from which information is gathered at regular periods lead the number of such classification rules to be generally so huge that selection of interesting ones among all discovered rules becomes an important task. At each period, using the newly gathered information from the domain, the new classification rules are induced. Therefore, these rules stream through time and are so called streaming classification rules. In this paper, an interactive classification rules' interestingness learning algorithm (ICRIL) is developed to automatically label the classification rules either as "interesting" or "uninteresting" with limited user interaction. In our study, VFFP (Voting Fuzzified Feature Projections), a feature projection based incremental classification algorithm, is also developed in the framework of ICRIL. The concept description learned by the VFFP is the interestingness concept of streaming classification rules.

From a given database for the extraction of hidden Classification Rules (CRs), One of the Data Mining technique is Classification Rule Mining (CRM), the objective being to build a classifier to classify "unseen" data. Survey of different techniques used for CRM has been done with thorough description of all dimensions of CRM.