Time Series Prediction

Predicting financial markets remains essential for decision-making in trading and investment. However, traditional models often fall short due to low-quality data, market volatility, and the complexity of global economic trends. Generative AI offers a transformative approach by improving data management processes, enhancing prediction accuracy, and overcoming the limitations of current forecasting models. This research explores the potential of Generative AI to enrich financial data through techniques like data augmentation, synthetic data generation, and anomaly detection. Specifically, the study investigates how integrating Generative AI into data preprocessing can enhance predictive modeling using Long Short-Term Memory (LSTM) networks for time series forecasting, particularly in stock prices. The research also explores feature engineering techniques such as moving averages, Relative Strength Index (RSI), and volatility metrics, which are crucial for capturing market trends. Generative Adversarial Networks (GANs) are employed to create additional realistic trading scenarios, boosting model robustness. Implemented using Python, the findings reveal that the combined LSTM and GAN model improved prediction accuracy over traditional methods, achieving an RMSE of 0.096 and R-squared (R²) 0.85 in stock price forecasts. The incorporation of synthetic data with traditional financial metrics leads to an improvement in the model's ability to predict short-term market trends. These findings suggest that the use of Generative AI can provide a more resilient and adaptive approach to financial market prediction, opening new avenues for research in the domain. Future research could focus on expanding the scope to include multi-market scenarios and incorporating real-time data streams.

This research study effectively attempted to estimate the trends in the area and production of onion in selected districts of Karnataka state. Onions are widely cultivated and used around the world. India is the second-largest onion producer in the world. For the present study the secondary data for 24 years dated from 19998-99 to 2021-22, collected on the area and production of onion for Vijayapura and Chitradurga districts of Karnataka from the official website of the Directorate of Economics and Statistics (2020), Department of Agriculture, Government of Karnataka. The data was analyzed through selected linear, quadratic, cubic, quartic, exponential and logistic models. The best-fit model was selected based on minimum Root Mean Square Error (RMSE), Mean Average Percentage Error (MAPE) values and highest R 2 Value. Results depicted that the quartic model was found best fit for the area of onion and the cubic model was found best fit for the production of onion in the Vijayapura district with fluctuation in trend. The quadratic model was found best fit for the area and production of onion in the Chitradurga district of Karnataka indicating an increasing trend.

Time series prediction is a problem with a wide range of applications, including energy systems planning, currency forecasting, stock exchange operations or traffic prediction. Accordingly, a number of different prediction approaches have been proposed such as linear models, Feedforward Neural network models, Recurrent Neural networks or Fuzzy Neural Models. In this paper one presents a prediction model based on fuzzy rules that relate past data values with the next unknown value to be estimated. A Fuzzy Boolean Neural Network has been used for that purpose and the laser data set of the Santa Fe contest has been used for illustration purpose. The results turned to be encouraging when compared with other published methods.

The use of neural networks for time series prediction has been an important focus of recent research. Multiobjective optimization techniques have been used for training neural networks for time series prediction. Cooperative coevolution is an evolutionary computation method that decomposes the problem into subcomponents and has shown promising results for training neural networks. This paper presents a multiobjective cooperative coevolutionary method for training neural networks where the training data set is processed to obtain the different objectives for multi-objective evolutionary training of the neural network. We use different time lags as multi-objective criterion. The trained multi-objective neural network can give prediction of the original time series for preprocessed data sets distinguished by their time lags. The proposed method is able to outperform the conventional cooperative coevolutionary methods for training neural networks and also other methods from the literature on benchmark problems.

The use of neural networks for time series prediction has been an important focus of recent research. Multiobjective optimization techniques have been used for training neural networks for time series prediction. Cooperative coevolution is an evolutionary computation method that decomposes the problem into subcomponents and has shown promising results for training neural networks. This paper presents a multiobjective cooperative coevolutionary method for training neural networks where the training data set is processed to obtain the different objectives for multi-objective evolutionary training of the neural network. We use different time lags as multi-objective criterion. The trained multi-objective neural network can give prediction of the original time series for preprocessed data sets distinguished by their time lags. The proposed method is able to outperform the conventional cooperative coevolutionary methods for training neural networks and also other methods from the literature on benchmark problems.

The foreign exchange market, characterized by high volatility and economic significance, requires accurate predictive models. This study investigates the application of the Temporal Fusion Transformer (TFT), enhanced with Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), for forecasting major foreign exchange (forex) currency pairs: USD/EUR, USD/JPY, USD/CNY, USD/AUD, and USD/INR. The proposed framework integrates a wide range of economic indicators, which include interest rate differentials, GDP growth, and trade balances, alongside investor sentiment derived from Twitter and ESG-related news sentiment. By addressing the non-linear, multiscale nature of forex time series, the CEEMDANTFT model facilitates improved signal decomposition and interpretability. Empirical results indicate that the model demonstrates competitive performance across all five currency pairs, with the USD/EUR pair exhibiting the highest predictive accuracy. Other pairs, exhibiting good predictive accuracy, include USD/JPY and USD/CNY, underscoring the model’s adaptability to varying economic contexts. Performance is assessed using multiple error metrics, and the model is benchmarked against standard neural network approaches (MLP, RNN, LSTM, CNN). Variable importance analysis highlights the dynamic influence of interest rates, investor sentiment, and ESG factors across different market regimes. This study provides empirical evidence that including ESG and investor sentiment can improve the accuracy of currency forecasting models. This study provides guidance and a framework for informed decision-making for traders, analysts, and policymakers.

Purpose: The purpose of this research was to provide a model for predicting time series of financial information based on the Lyapunov representation of information using chaos theory. Method: This research is applied in its purpose, which is conducted using a quantitative approach. The research ranks as descriptive-causal accounting research based on actual information in companies' financial statements. The research method is the "post-event" type and was carried out using chaos theory and Saida's method based on the Lyapunov view. Findings: The findings showed that during the ADF test, the null hypothesis was rejected at a level of less than 5% type 1 error and 95% confidence, and it shows that the data is not static. During the substitution analysis test and its significance level, the behavior of the time series of the main financial ‫ــــــــــــــــــــــــــ‬ ‫ــــ‬ ‫ـــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــ‬

Agricultural production and productivity forecasts are useful for farmers, policy makers and industries. In the present study, an auto-regressive and moving average model (ARIMA) has been applied for modelling and forecasting of annual coffee production and productivity in India. The augmented Dickey-Fuller (ADF) test has been used to test the stationarity of time series. The appropriate ARIMA model has selected based on the minimum Akaike information criterion (AIC). The residuals of the fitted models were diagnosed for the possible presence of autocorrelation and white noise effects, and Forecasts models are continuously increasing over the forecast period.

The problem of reliable detection of life-threatening situations in the neurosurgical patient undergoing treatment in the ICU is still far from reaching a satisfactory solution, although several methods of clinical and instrumental evaluation have recently been developed for the early detection of oncoming signs of danger. Continuous monitoring of intracranial pressure (ICP) provides neurosurgeons with valuable information about the current condition of the patient. However, it is increasingly felt that traditional methods of extracting information from the ICP signal have reached their natural limits, mostly because of difficulties in fitting the appropriate mathematical model to this nonlinear and non-stationary process. Successful implementations of artificial neural networks in many medical tasks have encouraged the application of this method to ICP processing. Two problems are considered: the prediction of trends in ICP, and recognition of the configuration of unfavourable symptoms likely to signal danger for the neurosurgical patient. The construction of neural network predictors of ICP trends is based on wavelet pre-processing of the original signal. The approach to the second task involves preprocessing of the ICP with spectral and statistical methods and classification of the extracted features of the current signal on an arbitrarily selected scale of danger.

Advances in metering technologies and emerging energy forecast strategies provide opportunities and challenges for predicting both short and long-term building energy usage. Machine learning is an important energy prediction technique, and is significantly gaining research attention. The use of different machine learning techniques based on a rolling-horizon framework can help to reduce the prediction error over time. Due to the significant increases in error beyond short-term energy forecasts, most reported energy forecasts based on statistical and machine learning techniques are within the range of one week. The aim of this study was to investigate how facility managers can improve the accuracy of their building’s long-term energy forecasts. This paper presents an extensive study of machine learning and data processing techniques and how they can more accurately predict within different forecast ranges. The Clarendon building of Teesside University was selected as a case study to ...

A feedforward neural network based on multi-valued neurons is considered in the paper. It is shown that using a traditional feedforward architecture and a high functionality multi-valued neuron, it is possible to obtain a new powerful neural network. Its learning does not require a derivative of the activation function and its functionality is higher than the functionality of traditional feedforward networks containing the same number of layers and neurons. These advantages of MLMVN are confirmed by testing using Parity n, two spirals and "sonar" benchmarks, and the Mackey-Glass time-series prediction.

 Why we need complex-valued neural networks?  The role of phase information  Multiple-Valued Logic (k-valued Logic) over the Field of the Complex Numbers  Multi-Valued Neuron (MVN)  Error-Correction Learning Rule for MVN  MVN with a periodic activation function and solving non-linearly separable problems  MVN-based multilayer feedforward neural network (MLMVN) and its derivative-free backpropagation learning algorithm based on the error-correction rule  MVN as a model of a biological neuron 2 (c) I. Aizenberg, IJCNN-2013 Highlights-2 (Applications)  Long-term time series prediction using MLMVN  Identification of blur and its parameters in image deblurring using MLMVN  Recognition of blurred images using MLMVN  Decoding of signals in an EEG-based brain-computer interface using MLMVN ( ) 1 B P z   (c) I. Aizenberg, IJCNN-2013 Phase-Dependent Activation Function  In fact, the activation function P B is a function of the argument (phase) of the weighted sum: it depends only on the argument of the weighted sum and does not depend on its magnitude. 20

In this paper, we observe some important aspects of Hebbian and error-correction learning rules for complex-valued neurons. These learning rules, which were previously considered for the multi-valued neuron whose inputs and output are located on the unit circle, are generalized for a complex-valued neuron whose inputs and output are arbitrary complex numbers. The Hebbian learning rule is also considered for the multi-valued neuron with a periodic activation function. It is experimentally shown that Hebbian weights, even if they still cannot implement an input/output mapping to be learned, are better starting weights for the error-correction learning, which converges faster starting from the Hebbian weights rather than from the random ones.

In this paper, a modified learning algorithm for the multilayer neural network with the multi-valued neurons (MLMVN) is presented. The MLMVN, which is a member of complex-valued neural networks family, has already demonstrated a number of important advantages over other techniques. A modified learning algorithm for this network is based on the introduction of an acceleration step, performing by means of the complex QR decomposition and on the new approach to calculation of the output neurons errors: they are calculated as the differences between the corresponding desired outputs and actual values of the weighted sums. These modifications improve significantly the existing derivative-free backpropagation learning algorithm for the MLMVN in terms of learning speed. A modified learning algorithm requires two orders of magnitude lower number of training epochs and less time for its convergence when compared to the existing learning algorithm. Good performance is confirmed not only by the much quicker convergence of the learning algorithm, but also by the compatible or even higher classification/prediction accuracy, which is obtained by testing over some benchmarks (Mackey-Glass and Jenkins-Box time series) and over some satellite spectral data examined in a comparison test.

Let (T, C, X ) be a vector of random variables (rvs) where T , C, and X are the interest variable, a right censoring rv, and a covariate, respectively. In this paper, we study the kernel conditional mode estimation when the covariate takes values in an infinite dimensional space and is α-mixing. Under some regularity conditions, the almost complete convergence of the estimate with rates is established.

Time series prediction is a problem with a wide range of applications, including energy systems planning, currency forecasting, stock exchange operations or traffic prediction. Accordingly, a number of different prediction approaches have been proposed such as linear models, Feedforward Neural network models, Recurrent Neural networks or Fuzzy Neural Models. In this paper one presents a prediction model based on fuzzy rules that relate past data values with the next unknown value to be estimated. A Fuzzy Boolean Neural Network has been used for that purpose and the laser data set of the Santa Fe contest has been used for illustration purpose. The results turned to be encouraging when compared with other published methods.

We describe a neural network collocation method (NNCM) for tomographic image reconstruction with small amount of projection data, which has been successfully applied to the three-dimensional ionospheric tomography based on the dataset of signal delays from the GPS satellites. In NNCM the neural network is trained by minimizing an object function composed of squared residuals of the governing equations evaluated at the collocation points and some constraining conditions imposed usually by observation data. This method is applied not only to the computerized tomography but also to the analyses of various inverse problems such as the data assimilation, the parameter estimation, the time series prediction, and so on.

In this study, artificial neural networks (ANN) and Adaptive-Network-Base fuzzy inference system (ANFIS) are used to model daily global solar radiation (GSR) in Tehran province of Iran. In order to design the networks, a dataset of meteorological daily time series for eight years (1994-2002) collected by Iran Meteorological Office was used. Input parameters were maximum temperature, relative sunshine duration, day of the year and extraterrestrial solar radiation while the output parameter was the GSR in MJ/m2 day. Various networks were designed and tested. The performances of best networks revealed that RMSE, MAE and MAPE were 2.77, 2.19, 0.12 for ANN and 2.8, 2.22, 0.12 for ANFIS, respectively. The results indicated that both approaches can be successfully applied for modeling GSR however ANN performs slightly better.

This study is based on time series analysis in forecasting government revenue, using federal government of Nigeria's statutory allocation to Lagos state as a case study. Monthly data were obtained from the federal ministry of finance, spanning a period of 10-11 years. Time series plot of the data was made in order to identify the time series components present in the data. A clear upward trend was seen to be present but the presence of (multiplicative) seasonal variations were not clear enough. Seasonal decomposition was carried out using the ratio-to-moving average method and the seasonal indices were found to be close to unity which therefore suggests that the seasonality component can be neglected. Different trend models were fitted and the S-curve trend model appeared to be the best among the fitted models, yielding the lowest mean squared deviation. Autoregressive Integrated Moving Average (ARIMA) models were developed, and these includes ARIMA (2,1,1), ARIMA (2,1,0) and ARIMA (0,1,1). Among these models, ARIMA (0,1,1) was the best model with the lowest Akaike's Information Criterion (AIC) and Schwarz's Bayesian Criterion (BIC). Overall, comparison of the models (that is, for both ARIMA models and least square trend models), ARIMA (0,1,1) was found to be the best model for the data and this was used in forecasting one year ahead.

Prediction of financial time series using artificial neural networks has been the subject of many publications, even if the predictability of financial series remains a subject of scientific debate in the financial literature. Facing this difficulty, analysts often consider a large number of exogenous indicators, which makes the fitting of neural networks extremely difficult. In this paper, we analyze how to aggregate a large number of indicators in a smaller number using -possibly nonlinear-projection methods. Nonlinear projection methods are shown to be equivalent to the linear Principal Component Analysis when the prediction tool used on the new variables is linear. Furthermore, the computation of the nonlinear projection gives an objective way to evaluate the number of resulting indicators needed for the prediction. Finally, the advantages of nonlinear projection could be further exploited by using a subsequent nonlinear prediction model. The methodology developed in the paper is validated on data from the BEL20 market index, using systematic cross-validation results.

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

A general-purpose useful parameter in time series forecasting is the regressor size, corresponding to the minimum number of variables necessary to forecast the future values of the time series. If the models are nonlinear, the choice of this regressor becomes very difÿcult. We present a quasi-automatic method using a nonlinear projection named curvilinear component analysis to build this regressor. The size of this regressor will be determined by the estimation of the intrinsic dimension of an over-sized regressor. This method will be applied to electric consumption of Poland using systematic cross-validation. The nonlinear model used for the prediction is a Kohonen map (self-organizing map).

Clustering methods are commonly used on time series, either as a preprocessing for other methods or for themselves. This paper illustrates the problem of clustering applied on regressor vectors obtained from row time series. It is thus shown why time series clustering may sometimes seem meaningless. A preprocessing is proposed to unfold time series and allow a meaningful clustering of regressors. Graphical and experimental results show the usefulness of the unfolding preprocessing.

Prediction of financial time series using artificial neural networks has been the subject of many publications, even if the predictability of financial series remains a subject of scientific debate in the financial literature. Facing this difficulty, analysts often consider a large number of exogenous indicators, which makes the fitting of neural networks extremely difficult. In this paper, we analyze how to aggregate a large number of indicators in a smaller number using -possibly nonlinear-projection methods. Nonlinear projection methods are shown to be equivalent to the linear Principal Component Analysis when the prediction tool used on the new variables is linear. The methodology developed in the paper is validated on data from the BEL20 market index.

Phenomenologists have provided a detailed description of the disorders of the subjective experience associated with minimal‐self disorders in patients with schizophrenia. Those patients report a range of distortions of their conscious experiences, including a sense of inner void, confusion between self and others, and, sometimes, a disruption of the sense of time. These reports have been interpreted as distortion of the first‐person perspective and a lack of immersion in the world, associated with a breakdown of the temporal structure of consciousness, and especially a disruption of the sense of time continuity. Further, it has been proposed that these disruptions are based on a difficulty to retain past information and to predict future information, that is, the mechanisms that help to relate events with one another and to reach a sense of time continuity. Experimental psychology results seem to converge to similar conclusions, inasmuch as some results in patients with schizophreni...

This paper presents a novel hybrid methodology for cryptocurrency price prediction that integrates chaos theory, quantile regression, and evolutionary optimization. We propose a three-stage framework: (1) chaos detection using the 0-1 test and phase-space reconstruction, (2) quantile-based rule generation via Particle Swarm Optimization (PSO), and (3) ensemble pruning for interpretable prediction intervals (PIs). The approach addresses critical limitations in existing methods by simultaneously modeling nonlinear market dynamics, heteroskedasticity, and prediction uncertainty. Evaluated on four major cryptocurrencies (Bitcoin SV, Maker, etc.), our model achieves 12-18% higher Prediction Interval Coverage Probability (PICP) compared to LSTM and XGBoost baselines, while maintaining an average rule set size below 10 rules per asset. The generated rules provide actionable insights into market regimes, demonstrating the framework's value for algorithmic trading and risk management in volatile financial environments.

In this paper we explicitly model the tail regions of the innovation distribution of two important series from the emerging financial markets of India, viz., Nifty, the equity index, and 30-day interest rate from the inter bank currency money market. Using the recent developments of extreme value theory, we estimate the tails by fitting a Generalised Pareto distribution to the observations lying beyond certain thresholds that mark the beginning of the tail regions. In line with the much discussed stylised features of financial returns, we find existence of tail-thickness, as indicated by the positive value of the tail indexes of the Generalised Pareto fit. Further, the left tail of each series is found to be heavier than the right tail. Thus, each series display asymmetric and heavy tailed behaviour.

A novel hybrid data-driven approach is developed for forecasting power system parameters with the goal of increasing the efficiency of short-term forecasting studies for non-stationary time-series. The proposed approach is based on mode decomposition and a feature analysis of initial retrospective data using the Hilbert-Huang transform and machine learning algorithms. The random forests and gradient boosting trees learning techniques were examined. The decision tree techniques were used to rank the importance of variables employed in the forecasting models. The Mean Decrease Gini index is employed as an impurity function. The resulting hybrid forecasting models employ the radial basis function neural network and support vector regression. Apart from introduction and references the paper is organized as follows. The second section presents the background and the review of several approaches for short-term forecasting of power system parameters. In the third section a hybrid machine learningbased algorithm using Hilbert-Huang transform is developed for short-term forecasting of power system parameters. Fourth section describes the decision tree learning algorithms used for the issue of variables importance. Finally in section six the experimental results in the following electric power problems are presented: active power flow forecasting, electricity price forecasting and for the wind speed and direction forecasting.

The inability of the capital asset pricing model (CAPM) to explain the crosssectional variation of average stock returns is well documented in the empirical asset pricing literature. Given the dismal performance of the static CAPM, numerous studies have investigated potential reasons for the empirical weakness. A prominent explanation for the empirical failure of the static CAPM is that the implications of conditioning information have been neglected. Most studies on conditional versions of the CAPM, allowing either for time-variation of the betas or the parameters of the stochastic discount factor, have found that by incorporating conditioning information, the empirical performance is clearly improved. The main motivation of this paper is to study the performance of the model class of the conditional CAPM in explaining the variation of average stock returns of the German stock market. Since previous empirical evidence is mainly based on studies for the U.S. stock market, we provide an additional robustness check for conditional asset pricing models. The unconditional CAPM and the Fama-French three-factor model serve as benchmarks in our analysis. 16 portfolios sorted by size and book-to-market are used as test assets. Our empirical results can be summarized as follows. We present further evidence regarding the empirical shortcomings of the unconditional CAPM in explaining German stock returns. Most importantly, our results suggest that incorporating conditioning information has the potential to substantially improve the model's performance. The choice of the conditioning variable, however, is crucial for the results. According to our empirical findings, a specification of the conditional CAPM with the term spread as the conditioning variable is able to explain the cross-section of returns about as well as the Fama-French model. Also the performance of the conditional CAPM using a January-dummy variable is encouraging, given the fact that it is only a slight modification of the standard model. There is no evidence that the use of conditioning variables necessarily leads to increased parameter instability. By contrast, the null hypothesis of stable parameters is rejected in the case of the three-factor model of Fama and French. Additional test results reveal however, that unconditional model specifications perform quite well in capturing the time-series predictability of the test asset returns.

We study the performance of conditional asset pricing models and multifactor models in explaining the German cross‐section of stock returns. We focus on several variables, which (according to previous research) are associated with market expectations on future market excess returns or business cycle conditions. Our results suggest that the empirical performance of the Capital Asset Pricing Model (CAPM) can be improved when allowing for time‐varying parameters of the stochastic discount factor. A conditional CAPM using the term spread explains the returns on our size and book‐to‐market sorted portfolios about as well as the Fama‐French three‐factor model and performs best in terms of the Hansen‐Jagannathan distance. Structural break tests do not necessarily indicate parameter instability of conditional model specifications. Another major finding of the paper is that the Fama‐French model – despite its generally good cross‐sectional performance – is subject to model instability. Uncon...

The dynamic behavior of mutation and crossover is investigated with the Breeder Genetic Algorithm. The main emphasis is on binary functions. The genetic operators are compared near their optimal performance. It is shown that mutation is most e cient in small populations. Crossover critically depends on the size of the population. Mutation is the more robust search operator. But the BGA combines the two operators in such a w ay that the performance is better than that of a single operator. For the DECEPTION function it is shown that increasing the size of the population above a certain number decreases the quality of the solutions obtained.

This paper is concerned with the automatic induction of parsimonious neural networks. In contrast to other program induction situations, network induction entails parametric learning as well as structural adaptation. We present a novel representation scheme called neural trees that allows efficient learning of both network architectures and parameters by genetic search. A hybrid evolutionary method is developed for neural tree induction that combines genetic programming and the breeder genetic algorithm under the unified framework of the minimum description length principle. The method is successfully applied to the induction of higher order neural trees while still keeping the resulting structures sparse to ensure good generalization performance. Empirical results are provided on two chaotic time series prediction problems of practical interest.

Efficient learning of a data analysis task strongly depends on the data representation. Many methods rely on symmetric similarity or dissimilarity representations by means of metric inner products or distances, providing easy access to powerful mathematical formalisms like kernel approaches. Similarities and dissimilarities are however often naturally obtained by non-metric proximity measures which can not easily be handled by classical learning algorithms. Major efforts have been undertaken to provide approaches which can either directly be used for such data or to make standard methods available for these type of data. We provide an overview about recent achievements in the field of learning with indefinite proximities.

Radiation dose in nuclear power plant reactors is known to be dominated by
the presence of radioisotopes in the primary loop of the reactor. In order to
strictly control it in normal operation (e.g., cleaning and reloading of nuclear
fuel), established chemical theories exist to explain the amount of radioisotopes
present in the reactor water circuits with respect to known control variables in
the plant (e.g., thermal power on the reactor, pH, hydrogen, etc.). However, the
high complexity and the uncertainty of the process make difficult an accurate
estimation of the measured values of radioisotopes. In order to address this
problem, this article introduces a dynamic Bayesian network (DBN) probabilistic
model that allows to experimentally demonstrate the capabilities of the control
variables to give information about the value of the radioisotope concentrations,
and to predict their values in a data-driven way. Our results in 5 different
nuclear power plants show that the accuracy and reliability of these predictions
is remarkable, enabling strategies for gathering reliable information about the chemical process in the primary loop, towards possible operational improvements.

In nuclear power plants, there are high-exposure jobs, like refuelling and maintenance, that require getting close to the reactor between operation cycles. Therefore, reducing radiation dose during these periods is of paramount importance regarding safety regulations. While there are some manipulable variables, like levels of certain corrosion products, that can influence the final level of radiation dose, there is no way to determine it in a principled way. In this work, we propose to use Machine Learning to predict the radiation dose in the reactor at the cycle end based on information available during the cycle operation. In particular, we use a Gaussian Process to model the relation between cobalt radioisotopes (a certain kind of corrosion product) and radiation dose levels. Gaussian Processes acknowledge the uncertainty on their predictions, a desirable property considering the high-risk nature of the present application. We report experiments on real data gathered from five different power plants in Spain. Results show that these models can be used to estimate the future values of radiation dose in a data-driven way. Moreover, there are tools based on these models currently in development for their application in power plants.

The field of time series forecasting has garnered significant attention in recent years, prompting the development of advanced models like TimeSieve, which demonstrates impressive performance. However, an analysis reveals certain unfaithfulness issues, including high sensitivity to random seeds, input and layer noise perturbations and parametric perturbations. Recognizing these challenges, we embark on a quest to define the concept of Faithful TimeSieve (FTS), a model that consistently delivers reliable and robust predictions. To address these issues, we propose a novel framework aimed at identifying and rectifying unfaithfulness in TimeSieve. Our framework is designed to enhance the model’s stability and faithfulness, ensuring that its outputs are less susceptible to the aforementioned factors. Experimentation validates the effectiveness of our proposed framework, demonstrating improved faithfulness in the model’s behavior.

Online shopping is an emerging area of corporate dialogue in terms of
survival and getting more market coverage. The methods of purchasing product is changing from traditional methods to online purchase facilitating shopping anytime with all benefits under one roof Technological inventions all over the world have changed the perception of consumer behavior. Consumers are playing a significant
function in online shopping.With the increase of digital engagement, the complexity
of taking purchase decision and chaotic behaviour are also portraited by the consumer.
The present study empirically asses the chaotic factors which sketches out the unpredictable consumer purchase behaviour through various shopping platforms. The study is based on Cross cultural survey and Primary data. Stratified sampling method is used to collect the sample. The respondent numbers to 120 for the analysis. The data
collected is analyzed by the statistical tools SPSS21 using the KMO and Barnett’s Test, to identify the major factors which are effective among the other factors that are influenced by the complexity; of the online shopping behaviour This study identifies the different pattern of online purchase, chaotic behaviour, online Purchase frequency complexities, age wise purchase pattern of the online shopping behaviour in market
prespectic.

This paper presents a new methodology for missing value imputation in a database. The methodology combines the outputs of several Self-Organizing Maps in order to obtain an accurate filling for the missing values. The maps are combined using MultiResponse Sparse Regression and the Hannan-Quinn Information Criterion. The new combination methodology removes the need for any lengthy cross-validation procedure, thus speeding up the computation significantly. Furthermore, the accuracy of the filling is improved, as demonstrated in the experiments.

Nous étudions l'estimateur à noyau du mode de la distribution d'une variable réelle Y conditionnée par une variable X à valeurs dans un espace semimétrique. Nous établissons les convergences en norme 1/ et presque complète de l'es- timateur. Les résultats asymptotiques établis,

Nous étudions l'estimateur à noyau du mode de la distribution d'une variable réelle Y conditionnée par une variable X à valeurs dans un espace semimétrique. Nous établissons les convergences en norme 1/ et presque complète de l'es- timateur. Les résultats asymptotiques établis,

This paper presents a comparative study of the hybrid models, neural networks and nonparametric regression models in time series forecasting. The components of these hybrid models are consisting of the nonparametric regression and artificial neural networks models. Smoothing spline, regression spline and additive regression models are considered as the nonparametric regression components. Furthermore, various multilayer perceptron algorithms and radial basis function network model are regarded as the artificial neural networks components. The performances of these models are compared by forecasting the series of number of produced Cars and Domestic product per capita (GDP) data occurred in Turkey. This comparisons show that hybrid models proposed in this paper have denoted much more excellent performance than the hybrid models in literature.

Military decision making demands an increasing ability to understand and structure the critical information on the battlefield. As the military evolves into a networked force, decision makers should select and filter information across the battlefield in a timely and efficient manner. Human capability in analyzing all the data is not sufficient because the modern battlefield is characterized by dramatic movements, unexpected evolutions, chaotic behavior and non-linear situations. The Artificial Intelligence (AI) ingredient permits to explore a greater range of options, enabling the staff to analyze more possible options in the same amount of time, together with a deeper analysis of these options.

This paper presents a novel approach for the simultaneous modelling and forecasting of wind signal components. This is achieved in the complex domain by using novel neural network algorithms and architectures. We first perform a signal nonlinearity and component-dependent analyses, which suggest the use of modular complex-valued recurrent neural networks (RNNs). This RNN-based modelling rests upon a combination of nonlinearity, complexity and internal memory and allows for the multiple step ahead forecasting of the wind signal in its complex form (speed and direction). The approach is first verified on benchmark Data Set A (NH 3 laser data) of the Santa Fe Time Series Prediction Competition together with artificial data generated by chaotic Mackey-Glass equations, and then applied to the real-world wind measurements. Simulations support the proposed architecture and algorithms.

We discuss and compare measures of accuracy of univariate time series forecasts. The methods used in the M-competition and the M3-competition, and many of the measures recommended by previous authors on this topic, are found to be degenerate in commonly occurring situations. Instead, we propose that the mean absolute scaled error become the standard measure for comparing forecast accuracy across multiple time series.

In the present study, we have used Box-Jenkins approaches an Autoregressive Integrated Moving Average model (ARIMA) for modeling and forecasting of annual amount of Arabica and Robusta coffee production and yield (ARCPY) in India. In this study used time series data was collected from the official website of the coffee board of India from 1986 to 2023 (38 observations). Augmented Dickey-Fuller (ADF) test has used for testing the stationarity of the time series, and the appropriate ARIMA model has selected based on minimum Akaike Information Criterion (AIC). The ARIMA models has compared with the other ARIMA models with respect to forecast accuracy measures, and the residuals has diagnosed for possible presence of autocorrelation, and white noise heteroscedasticity (WNH) test of the fitted models. The MAPE value of ACP and ACY has 8.94 and 9.35 percent respectively shows highly accurate forecasting percentage rate respectively. While, the MAPE value of RCP and RCY has 16.03 and 11.43 percent respectively shows good accurate forecasting percentage rate respectively. Thus, we found the ARIMA (2, 1, 4), (3, 1, 2), (0, 1, 3) and Original Research Article

Predicting floods is crucial in minimizing the
harmful consequences of floods, particularly with
changing weather patterns and increasing urban
development. This review article explores how
transfer learning methods and ensemble machine
learning have been utilized to improve the accuracy
and reliability of flood prediction advancements.
Conventional flood prediction techniques often
encounter problems like inadequate data sharing and
restricted geographical suitability. Ensemble machine
learning approaches like boosting, stacking, and
bagging can address these difficulties and improve
prediction accuracy. Transfer learning methods offer
advantages by using data from one location to build
models in another, effectively tackling issues linked
to limited data access.
This research integrates current studies on AIbased
urban flood control systems with real-time
data integration to provide a thorough summary of
recent progress in flood prediction. Important
patterns, challenges, and chances in the industry are
emphasized following a comprehensive examination
of the current literature. Ultimately, the report
provides suggestions for future research paths and
developing more accurate and reliable flood
prediction systems.

Slow feature analysis (SFA) is a method for extracting slowly varying driving forces from quickly varying nonstationary time series. We show here that it is possible for SFA to detect a component which is even slower than the driving force itself (e.g. the envelope of a modulated sine wave). It is shown that it depends on circumstances like the embedding dimension, the time series predictability, or the base frequency, whether the driving force itself or a slower subcomponent is detected. We observe a phase transition from one regime to the other and it is the purpose of this work to quantify the influence of various parameters on this phase transition. We conclude that what is percieved as slow by SFA varies and that a more or less fast switching from one regime to the other occurs, perhaps showing some similarity to human perception.