Search

With the increasing use of sensors in almost every device and application worldwide, the volume of data increases exponentially. Instance selection is an important data preprocessing step that can be applied to many machine learning models. It is even crucial for executing machine learning tasks in constrained environments as demanded by many Internet of Things (IoT) enabled applications today such as the Human Activity Recognition (HAR) domain of applications. In this paper, we have proposed a hybrid selection and training pipeline methodology that combines the nearest neighbor concept with evolutionary computing to address the instance selection problem of smartphone sensing-based HAR. A clustering algorithm has been proposed first followed by a Genetic Algorithm based instance selection approach. It is so far the first of its kind in the aforementioned domain of HAR based on smartphone sensing data. The proposed instance selection has been tested on two of the popular benchmark HAR datasets — UCI-HAR and WISDM. The experimented outputs report that the proposed method has effectively reduced the dataset size by around 40% for the benchmark datasets while retaining the recognition accuracy above 94%. This is a clear depiction of the removal of the outliers from the instance set. The proposed approach has also been compared with other state-of-the-art under-sampling approaches and the results show that our approach has performed better. • Tuple selection pipeline for Human Activity Recognition in constrained environments. • Combination of unsupervised approach with Genetic Algorithm for Instance selection. • 40% reduction of instances while retaining accuracy above 94%. • Proposed pipeline performs better than state-of-the art undersampling approaches. • Hardness property of instances is found to decrease. [ABSTRACT FROM AUTHOR]

• Using an optimal cost-sensitive credit scoring. • Using a hybrid performance metric to calculate the probability of correct scoring. • Using an optimal feature selection technique to nine different scenarios. • A policy advice to banking activists to resort to a combination of models. This paper aims to provide an optimal credit rating by using a new hybrid performance metric, including several performance metrics, referred to as the probability of credit scoring correctness. There have been several credit-scoring models proposed in the banking system. These models, either single classifiers or hybrid and ensemble models determine the class of customers (good or bad accounts). The banks need the probability of default to calculate the expected loss and the unexpected loss and then determine whether their capital is adequate. Generally speaking, the proposed model can utilize n classifiers and m performance metrics, the larger n , and m , the more diverse and more reliable the model is as an estimate of customer classes. We evaluated the proposed model using five common classifiers, namely k-nearest neighbors (KNN), support vector machine (SVM), decision tree (DT), AdaBoost, and probabilistic neural network (PNN), and eight performance metrics. To extract the feature, an optimization approach using genetic algorithms and nine different scenarios were used. We have then calculated the optimal parameters of each algorithm simultaneously by defining the cost-sensitive objective function based on the probability of correctly classifying it into two classes 0 (good) and 1 (bad). We evaluated the performance of the proposed hybrid model using the Iranian dataset and the UCI German and Australian datasets. The results indicate the proposed method provides very good scoring performance. Using this approach, the bank can filter its bad customers based on the tolerance it considers for the probability value, so it can decide whether it should provide loans to that customer. [ABSTRACT FROM AUTHOR]

• Forecast electricity demand considering residents' psychological preferences. • Stationarity time series are established to avoid the impact of irregular factors. • Personalizes exogenous variables, combines correlations and importance. • A hybrid machine learning based model is built to improve forecasting performance. • Multiple implications are provided for the government and institutions. Faced with the explosive growth of population and rapid development of industrialization, household electricity demand (HED) has recently become an essential sector of energy demand. Very-short-term forecasting of HED is closely related to monitoring insecurity, detecting emergency and potential crisis. Consider challenges triggered by the characteristics of time series, firstly, a phase of data pre-processing with stationarity of time series is constructed to manage irregular trend factors. Subsequently, in the feature selection phase, a two-layer feature selection method is proposed with the combination of random forest and autocorrelation analysis to select the most relevant features, which simultaneously considers correlations and importance of input lagged variables as well as exogenous variables. After that, hybrid HED forecasting methods are constructed with incorporating multiple machine learning techniques in the forecasting phase, namely k-nearest neighbor, support vector regression , random forest and multilayer perceptron, using their respective advantages of managing different situations. As residents' psychological preferences and calendar variables highly affect future HED, this paper intends to add and test their effects on HED. Finally, the proposed methods are examined on a 10-min-ahead horizon application, whose forecasting results, comparison analyses and an extended experiment are given to illustrate their effectiveness, after which some implications are provided to show the insights gained from the results. [ABSTRACT FROM AUTHOR]

• Development of the innovative machine learning based design support system (DesSS). • Comparison with other state-of-the-art machine learning and deep learning model. • Validation of the DesSS on two real use case datasets. • High computational speed and accuracy compared to simulation tools. • Trade-off between the model interpretability, computation effort and accuracy. In the engineering practice, it frequently occurs that designers, final or intermediate users have to roughly estimate some basic performance or specification data on the basis of input data available at the moment, which can be time-consuming. There is the need for a tool that will fill the missing gap in the optimization problems in engineering design processes, by making use of the advances in the artificial intelligence field. This paper aims to fill this gap by introducing an innovative Design Support System (DesSS), originated from the Decision Support System, for the prediction and estimation of machine specification data such as machine geometry and machine design on the basis of heterogeneous input parameters. As the main core of the developed DesSS, we introduced different machine learning (ML) approaches based on Decision/Regression Tree, k-Nearest Neighbors, and Neighborhood Component Features Selection. Experimental results obtained on a real use case and using two different real datasets demonstrated the reliability and the effectiveness of the proposed approach. The innovative machine learning-based DesSS meant for supporting the designing choice, can bring various benefits such as the easier decision-making, conservation of the company's knowledge, savings in man-hours, higher computational speed and accuracy. [ABSTRACT FROM AUTHOR]