F. Baratzadeh; Seyed M. H. Hasheminejad
Abstract
With the advancement of technology, the daily use of bank credit cards has been increasing exponentially. Therefore, the fraudulent use of credit cards by others as one of the new crimes is also growing fast. For this reason, detecting and preventing these attacks has become an active area of study. ...
Read More
With the advancement of technology, the daily use of bank credit cards has been increasing exponentially. Therefore, the fraudulent use of credit cards by others as one of the new crimes is also growing fast. For this reason, detecting and preventing these attacks has become an active area of study. This article discusses the challenges of detecting fraudulent banking transactions and presents solutions based on deep learning. Transactions are examined and compared with other traditional models in fraud detection. According to the results obtained, optimal performance is related to the combined model of deep convolutional networks and short-term memory, which is trained using the aggregated data received from the generative adversarial network. This paper intends to produce sensible data to address the unequal class distribution problem, which is far more effective than traditional methods. Also, it uses the strengths of the two approaches by combining deep convolutional network and Long Short Term Memory network to improve performance. Due to the inefficiency of evaluation criteria such as accuracy in this application, the measure of distance score and the equal error rate has been used to evaluate models more transparent and more precise. Traditional methods were compared to the proposed approach to evaluate the efficiency of the experiment.
S. Bayatpour; Seyed M. H. Hasheminejad
Abstract
Most of the methods proposed for segmenting image objects are supervised methods which are costly due to their need for large amounts of labeled data. However, in this article, we have presented a method for segmenting objects based on a meta-heuristic optimization which does not need any training data. ...
Read More
Most of the methods proposed for segmenting image objects are supervised methods which are costly due to their need for large amounts of labeled data. However, in this article, we have presented a method for segmenting objects based on a meta-heuristic optimization which does not need any training data. This procedure consists of two main stages of edge detection and texture analysis. In the edge detection stage, we have utilized invasive weed optimization (IWO) and local thresholding. Edge detection methods that are based on local histograms are efficient methods, but it is very difficult to determine the desired parameters manually. In addition, these parameters must be selected specifically for each image. In this paper, a method is presented for automatic determination of these parameters using an evolutionary algorithm. Evaluation of this method demonstrates its high performance on natural images.
H.3.2.4. Education
Seyed M. H. Hasheminejad; M. Sarvmili
Abstract
Nowadays, new methods are required to take advantage of the rich and extensive gold mine of data given the vast content of data particularly created by educational systems. Data mining algorithms have been used in educational systems especially e-learning systems due to the broad usage of these systems. ...
Read More
Nowadays, new methods are required to take advantage of the rich and extensive gold mine of data given the vast content of data particularly created by educational systems. Data mining algorithms have been used in educational systems especially e-learning systems due to the broad usage of these systems. Providing a model to predict final student results in educational course is a reason for using data mining in educational systems. In this paper, we propose a novel rule-based classification method, called S3PSO (Students’ Performance Prediction based on Particle Swarm Optimization), to extract the hidden rules, which could be used to predict students’ final outcome. The proposed S3PSO method is based on Particle Swarm Optimization (PSO) algorithm in discrete space. The S3PSO particles encoding inducts more interpretable even for normal users like instructors. In S3PSO, Support, Confidence, and Comprehensibility criteria are used to calculate the fitness of each rule. Comparing the obtained results from S3PSO with other rule-based classification methods such as CART, C4.5, and ID3 reveals that S3PSO improves 31 % of the value of fitness measurement for Moodle data set. Additionally, comparing the obtained results from S3PSO with other classification methods such as SVM, KNN, Naïve Bayes, Neural Network and APSO reveals that S3PSO improves 9 % of the value of accuracy for Moodle data set and yields promising results for predicting students’ final outcome.
H.3. Artificial Intelligence
Seyed M. H. Hasheminejad; Z. Salimi
Abstract
One of the recent strategies for increasing the customer’s loyalty in banking industry is the use of customers’ club system. In this system, customers receive scores on the basis of financial and club activities they are performing, and due to the achieved points, they get credits from the ...
Read More
One of the recent strategies for increasing the customer’s loyalty in banking industry is the use of customers’ club system. In this system, customers receive scores on the basis of financial and club activities they are performing, and due to the achieved points, they get credits from the bank. In addition, by the advent of new technologies, fraud is growing in banking domain as well. Therefore, given the importance of financial activities in the customers’ club system, providing an efficient and applicable method for detecting fraud is highly important in these types of systems. In this paper, we propose a novel sliding time and scores window-based method, called FDiBC (Fraud Detection in Bank Club), to detect fraud in bank club. In FDiBC, firstly, based on each score obtained by customer members of bank club, 14 features are derived, then, based on all the scores of each customer member, five sliding time and scores window-based feature vectors are proposed. For generating training and test data set from the obtained scores of fraudster and common customers in the customers’ club system of a bank, a positive and a negative label are used, respectively. After generating training data set, learning is performed through two approaches: 1) clustering and binary classification with OCSVM method for positive data, i.e. fraudster customers, and 2) multi-class classification including SVM, C4.5, KNN, and Naïve Bayes methods. The results reveal that FDiBC has the ability to detect fraud with 78% accuracy and thus can be used in practice.
