-
Open Access Article
1 - Introducing a new meta-heuristic algorithm to solve the feature selection problem
Mehdi Khadem Abbas Toloie Eshlaghy Kiamars Fathi HafshejaniDue to the increase in the volume of data and information in recent years, the issue of choosing the most appropriate feature for decision making has become very important. Classic attribute selection methods cannot work well on big data. Because feature selection is a MoreDue to the increase in the volume of data and information in recent years, the issue of choosing the most appropriate feature for decision making has become very important. Classic attribute selection methods cannot work well on big data. Because feature selection is a complex problem, it seems appropriate to use meta-heuristic algorithms to solve this problem. In this paper, a new meta-heuristic algorithm inspired by nomadic migration to solve the feature selection problem is presented. This algorithm is named in honor of the Qashqai tribe. In this hybrid algorithm, the proportional function was designed based on the feature selection algorithm and based on minimizing the number of features and the amount of data error using neural network results. Then the Qashqai meta-heuristic algorithm was implemented on this fitness function and the results were compared with the well-known meta-heuristic algorithms of genetics and particle swarm. The results of the hypothesis test showed that the Qashqai optimization algorithm to solve the feature selection problem by the genetic algorithm and particle swarm is not defeated and in terms of convergence to the optimal solution works well. Manuscript profile -
Open Access Article
2 - Improving the Operation of Text Categorization Systems with Selecting Proper Features Based on PSO-LA
Mozhgan Rahimirad Mohammad Mosleh Amir Masoud Rahmani -
Open Access Article
3 - Improvement of effort estimation accuracy in software projects using a feature selection approach
Zahra Shahpar Vahid Khatibi Asma Tanavar Rahil Sarikhani -
Open Access Article
4 - A New Approach for Text Documents Classification with Invasive Weed Optimization and Naive Bayes Classifier
Saman Khalandi Farhad Soleimanian Gharehchopogh -
Open Access Article
5 - A Novel Hybrid Approach for Email Spam Detection based on Scatter Search Algorithm and K-Nearest Neighbors
Samira Amjad Farhad Soleimanian Gharehchopogh -
Open Access Article
6 - Improve Spam Detection in the Internet Using Feature Selection based on the Metahuristic Algorithms
Abdulbaghi Ghaderzadeh sahar Hosseinpanahi Sarkhel Taher kareem -
Open Access Article
7 - An Optimization-based Learning Black Widow Optimization Algorithm for Text Psychology
Ali Hosseinalipour Farhad Soleimanian Gharehchopogh mohammad masdari ALi Khademi -
Open Access Article
8 - improving intrusion detection systems by feature reducing based on genetics algorithm and data mining techniques
Mehdi Keshavarzi hossein MomenzadehThe network-based computer systems play critical role in our modern society; so there is highly chance these systems might be target of intrusion and attacks. In order to implement full-scale security in a computer network, firewalls and other intrusion prevention mecha MoreThe network-based computer systems play critical role in our modern society; so there is highly chance these systems might be target of intrusion and attacks. In order to implement full-scale security in a computer network, firewalls and other intrusion prevention mechanisms aren’t always enough and needs other systems called intrusion detection systems. An Intrusion detection system can be set of tools, algorithms and evidence that help to identify, locate and report illegal or not approved activities by the network. Intrusion detection systems can be established by software or hardware systems and each have their own advantages and disadvantages. Because of various characteristics of intrusion detection data, in this research we select effective characteristics using improved genetic algorithm. Then by means of standard data mining techniques, we present a model for data classification.For performance evaluation of this suggested method, we used NSL-KDD database that has more realistic records than other intrusion detection data. Manuscript profile -
Open Access Article
9 - Using Ant Colony Algorithm and Pairwise Learning to Classify Attack in Intrusion Detection Systems
Mohammad Ali Nadoomi Majid SinaIntrusion detection systems for security in computer networks have been proposed to be crossed if the attacker from other security equipment, able to detect it and prevent it from advancing. One of the challenges of these systems, it is high dimensional data. In this st MoreIntrusion detection systems for security in computer networks have been proposed to be crossed if the attacker from other security equipment, able to detect it and prevent it from advancing. One of the challenges of these systems, it is high dimensional data. In this study was to reduce the dimensions of a simple genetic algorithm with the length of the string variable we use. Then, according to selected characteristics, a meta-heuristic model for data classification, using ant colony algorithm offer. Classification model proposed by trying to divide the data into two samples is Hnjydh and Nahnjydh. The proposed method for evaluating the performance of database intrusion detection NSL-KDD than other data from the records of more realistic approach is used. The results of the experiments, the proposed method has better performance compared with other existing methods show. Manuscript profile -
Open Access Article
10 - Offering a model for persian texts classify by combination of classification methods
iman jamali Seyed Javad Mirabedini علی HarounabadiTo classify text information extraction techniques, natural language processing and machine learning has been widely used general purpose of categories of documents, classified documents in the form of a certain number of categories are pre-determined. Each document can MoreTo classify text information extraction techniques, natural language processing and machine learning has been widely used general purpose of categories of documents, classified documents in the form of a certain number of categories are pre-determined. Each document can be in one, several or no category is placed. In the case of any document to this question will be placed the document on which of the categories. This can be in the form of an automatic learning to use it any document can be automatically assigned to a category. In this thesis, data collection and cleanup after you select text using the normal method of word frequency -inverse document frequency (norm TF-IDF) is the weight features and features in two stages using document frequency (DF) and Chi square (SChi) are selected, and then using principal component analysis (PCA) features reduced dimensions, and at a later stage by combining 21 support vector machine (SVM) the proposed model we have implemented, and the accuracy of the model to assess the 10-step method validation. Experimental results show that this model can text classification accuracy of 91.86 for the seven categories do, which has a higher accuracy than the earlier work done. Manuscript profile -
Open Access Article
11 - Presenting A Hybrid Method of Deep Neural Networks to Prevent Intrusion in Computer Networks
Mohsen Roknaldini Erfaneh NorooziIntroduction: Nowadays, computer networks have significant impacts on our daily lives, leading to cybersecurity becoming a crucial area of research. Cybersecurity techniques mainly encompass antivirus software, firewalls, and intrusion detection systems. Intrusion dete MoreIntroduction: Nowadays, computer networks have significant impacts on our daily lives, leading to cybersecurity becoming a crucial area of research. Cybersecurity techniques mainly encompass antivirus software, firewalls, and intrusion detection systems. Intrusion detection system is one of the fundamental security tools in the field of computer networks and systems. The primary goal of an intrusion detection system is to identify and alert about any unauthorized activities, threats, or attacks on a system or network. By analyzing the flow of data and network/system events, the intrusion detection system attempts to identify patterns and indicators related to various attacks and intrusions. Intrusion detection systems can operate based on rules or learning. In the rule-based approach, algorithms and rules created by security experts and analysts are used to detect patterns and identify attacks. However, in the machine learning approach, machine learning algorithms and deep neural networks are employed to extract patterns and features related to attacks from real data. Method: This study focuses on the examination and presentation of a combined approach using deep neural networks to prevent intrusions in computer networks. The primary objective of this research is to enhance the efficiency of intrusion detection systems. To achieve this goal, a combined approach of deep learning and artificial neural networks is proposed. This approach utilizes deep neural networks to detect more complex features and improves the model's performance. Results: Simulation results demonstrate that deep neural network methods such as MLP, CNN, LSTM, and GRU yield favorable outcomes compared to other single-layer machine learning techniques. In this study, two combined methods, CNN-GRU and CNN-LSTM, were introduced and tested on the KDD CUP'99 dataset for comprehensive analysis and evaluation. Both combined approaches exhibit high accuracy and lower classification errors compared to other introduced methods. Therefore, it can be concluded that the CNN-LSTM combined approach performs well on the KDD CUP'99 dataset. Discussion: Based on the achieved results, the combined CNN-LSTM and CNN-GRU methods offer very good performance with accuracies of 99.95% and 99.92%, respectively, on the KDD CUP'99 dataset. Among these methods, minor differences in the performance of some parameters for classes may exist, yet both approaches remain acceptable. Hence, it can be concluded that the combined CNN-LSTM approach performs well on the KDD CUP'99 dataset. Manuscript profile -
Open Access Article
12 - A New Approach Based on Deep Learning Algorithms to Study Effective Factors of Using Social Networks on Students’ Performance
Maryam Bakeshlo Mohammad Tahghighi SharabyanIntroduction: In the past few years, the emerging technologies of mobile phones have grown rapidly and led to the creation of a new category of social media, which are very efficient mechanisms for collaboration and communication between their users. Social media consis MoreIntroduction: In the past few years, the emerging technologies of mobile phones have grown rapidly and led to the creation of a new category of social media, which are very efficient mechanisms for collaboration and communication between their users. Social media consists of a variety of web-based tools that enable users to distribute and share new ideas, thoughts, and information in an interactive environment. Social networks and the use of these networks have become an inseparable part of the lives of many students, so it has a direct impact on all aspects of their lives, including academic performance. Therefore, in this research, a new approach based on data mining techniques will be presented to investigate the factors affecting the use of social networks in the academic performance of students, and for this purpose, the deep learning technique and classification will be used.Method: The proposed method has two main phases. In the first phase, the data is prepared for modeling during the three stages of data integration, cleaning, and transformation, and in the final phase, the data is modeled and analyzed using deep learning.Results: The presented method has a favorable output in the feature selection set with a value of 68%, which shows an improvement of 14% compared to the basic method, which can be concluded that about 68% of social networks and the use of the Internet on the learning and efficiency of students is effective.Discussion: Social networking sites are very useful in education and research considering that they are used in schools for organizational branding, recruiting, and encouraging students and employees to participate. In this research, a method to identify the impact of social networks and internet use on students' learning based on accurate classification has been presented. Manuscript profile -
Open Access Article
13 - An Online group feature selection algorithm using mutual information
maryam rahmaninia sondos bahadoriIntroduction: In the area of big data, the dimension of data in many fields are increasing dramatically. To deal with the high dimensions of training data, online feature selection algorithms are considered as very important issue in data mining. Recently, online featur MoreIntroduction: In the area of big data, the dimension of data in many fields are increasing dramatically. To deal with the high dimensions of training data, online feature selection algorithms are considered as very important issue in data mining. Recently, online feature selection methods have attracted a lot of attention from researchers. These algorithms deal with the process of selecting important and efficient features and removing redundant features without any pre-knowledge of the set of features. Despite all the progress in this field, there are still many challenges related to these algorithms. Among these challenges, we can mention scalability, minimum size of selected features, sufficient accuracy and execution time. On the other hand, in many real-world applications, features are entered into the dataset in groups and sequentially. Although many online feature selection algorithms have been presented so far, but none of them have been able to find trade of between these criteria. Method: In this paper, we propose a group online feature selection method with feature stream using two new measures of redundancy and relevancy using mutual information theory. Mutual information can compute linear and non-linear dependency between the variables. With the proposed method, we try to create a better tradeoff between all the challenges. Results: In order to show the effectiveness of the proposed online group feature selection method, a number of experiments have been conducted on six large multi-label training data sets named ALLAML, colon, SMK-CAN-187, credit-g, sonar and breast-cancer in different applications and 3 online group feature selection algorithms named FNE_OGSFS، Group-SAOLA and OGSFS which are presented recently. Also, 3 evaluation criteria including average accuracy using KNN (k - nearest neighborhood (, SVM (Support Vector Machine) and NB (Naïve Bayesian) classifiers, number of selected features and executing time were used as criteria for comparing the proposed method. According to the obtained results, the proposed algorithm has obtained better results in almost of cases compared to other algorithms which it shows the efficiency of the proposed method. Discussion: In this paper, we will show that proposed online group feature selection method will achieve better performance by considering label group dependency between the new arrival features. Manuscript profile -
Open Access Article
14 - Stable Feature Selection and Clustering According to Hierarchical Structures Based on Chaotic Multispecies Particle Swarm Optimization Applied for Genetic Data Diagnosis and Prognosis
Maryam yassi Mohammad Hossein Moattarb Mehdi YaghoobiAny abnormal reproduction of cells is a tumor. censer happens when there’s an unstrained growth of abnormal cells. Cancer and tumors are divided in to two types, malignant and benign. Given the growth in the environmental information, it’s essential to emplo MoreAny abnormal reproduction of cells is a tumor. censer happens when there’s an unstrained growth of abnormal cells. Cancer and tumors are divided in to two types, malignant and benign. Given the growth in the environmental information, it’s essential to employ some tools to analyze this data and gain the knowledge embedded in it. Since large-scale problems and huge data bases are incomprehensible for the human, employing intelligent methods is effective in understanding large-scale data better. In this paper, the integration methods are a subset of rating measures each with a specific objective of sustainable features for superior selection of distinct features.The next step would discuss creating a fuzzy system (FS) to detect and classify between benign and malignant nature of biological data. Fuzzy system type is Takagi-Sugeno-Kang (TSK). To classify a hierarchical structure of multi-species particle swarm algorithm based on chaotic particle can be used to optimize the fuzzy system. In addition, using chaotic theory discerns the true diversity of the particles and increases the power to detect and classify the samples. Accurate identification and classification of malignant and benign biological nature of the data is more than 95%. This simulation is performed on UCI and Microarray data-base. Manuscript profile -
Open Access Article
15 - Predict the Stock price crash risk by using firefly algorithm and comparison with regression
Serveh Farzad Esfandiar Malekian Hossein Fakhari Jamal Ghasemi -
Open Access Article
16 - A Combinatory Feature Selection Method using Gray Wolf Optimization and Crow Search Algorithms for Intrusion Detection Systems
Kayvan Asghari -
Open Access Article
17 - The Mechanical Design of Drowsiness Detection Using Color Based Features
Peyman jabraelzade Rahim parikhani -
Open Access Article
18 - A Review of Feature Selection Method Based on Optimization Algorithms
Zohre Sadeghian Ebrahim Akbari Hossein Nematzadeh Homayun Motameni -
Open Access Article
19 - Feature Selection Using Multi Objective Genetic Algorithm with Support Vector Machine
Mojgan Elikaei Ahari Babak Nasersharif -
Open Access Article
20 - Presenting a New Approach for Detecting Attacks on Voice over Internet Protocol Based on Ensemble Clustering
Farid Bavifard Mohammad Kheyrandish Mohammad MoslehDue to lower cost and greater flexibility, voice over internet protocol (VoIP) is widely used in telecommunications. A variety of VoIP terminals causes them to be vulnerable. A common way to secure VoIP includes intrusion detection based on machine learning. Due to the MoreDue to lower cost and greater flexibility, voice over internet protocol (VoIP) is widely used in telecommunications. A variety of VoIP terminals causes them to be vulnerable. A common way to secure VoIP includes intrusion detection based on machine learning. Due to the diversity of traffics and lack of class labels for training Intrusion detection systems (IDSs) in many situations, clustering approaches (unsupervised learning) have been focused on. But individual cluster systems can't cover the diversities of feature values well, and some traffic samples may be identified as outliers. As an ensemble approach, the proposed model for solving these problems focuses on using TwoStep clustering algorithm, and by improving it, tries to improve the clustering-based intrusion detection. Moreover, regarding the importance of the feature selection process, a combination of Simulated Annealing algorithm (SA) and Multi-Layer Perceptron (MLP) has been exploited for identifying superior features used for clustering VoIP packets, as Normal or involving DoS, R2L, U2R either Probe attacks. Based on evaluation results obtained on the dataset “Network Security Lab-Knwledge Discovery in Databases” (NSL-KDD) by MATLAB, the proposed feature selection reduced the training and testing times, averagely by 77% and 80%, respectively, by reducing the features to 10 and 8. Also, compared to previous works, the proposed IDS shows average improvements in Accuracy, Detection rate, and F-Measure at 3.34 %, 14.17 %, and 32.87 %, respectively. Manuscript profile -
Open Access Article
21 - Optimal Feature Space Selection in Detecting Epileptic Seizure based on Recurrent Quantification Analysis and Genetic Algorithm
Saleh LAshkari Mehdi AzarnooshSelecting optimal features based on nature of the phenomenon and high discriminant ability is very important in the data classification problems. Since it doesn't require any assumption about stationary condition and size of the signal and the noise in Recurrent Quantif MoreSelecting optimal features based on nature of the phenomenon and high discriminant ability is very important in the data classification problems. Since it doesn't require any assumption about stationary condition and size of the signal and the noise in Recurrent Quantification Analysis (RQA), it may be useful for epileptic seizure Detection. In this study, RQA was used to discriminate ictal EEG from the normal EEG where optimal features selected by combination of algorithm genetic and Bayesian Classifier. Recurrence plots of hundred samples in each two categories were obtained with five distance norms in this study: Euclidean, Maximum, Minimum, Normalized and Fixed Norm. In order to choose optimal threshold for each norm, ten threshold of ε was generated and then the best feature space was selected by genetic algorithm in combination with a bayesian classifier. The results shown that proposed method is capable of discriminating the ictal EEG from the normal EEG where for Minimum norm and 0.1˂ε˂1, accuracy was 100%. In addition, the sensitivity of proposed framework to the ε and the distance norm parameters was low. The optimal feature presented in this study is Trans which it was selected in most feature spaces with high accuracy. Manuscript profile -
Open Access Article
22 - Modeling factors influencing the confusion of female customers in choosing retail stores
Soheila ZarinJoy alvar Maryam Nooraei AbadehToday, the shopping center is no longer a place for transaction purposes where customers go to if they need a product or service, but it is a place for social purposes. Shoppers may become confused if they perceive a social environment as too stimulating or inappropriat MoreToday, the shopping center is no longer a place for transaction purposes where customers go to if they need a product or service, but it is a place for social purposes. Shoppers may become confused if they perceive a social environment as too stimulating or inappropriate. The aim of the current research is to combine interpretive structural modeling and machine learning to stratify factors affecting customer confusion in choosing a retail store. In order to select the features that are most related to the target variables and can provide the best performance in predicting and interpreting the model, feature selection using a neural network has been used. The proposed approach initially reduces the computational complexity in solving the design problem by reducing the dimensions of the problem space through the training of a type of multilayer neural networks. 7 factors affecting customer confusion were extracted based on this neural network model, classified in 5 levels using interpretative structural modeling, and the related model was drawn. In this category, the variables of a wide variety of store brands, loyalty programs and the concentration of stores in one location, the greatest power of influence (influence) and perceived risk, individual and demographic factors and the volume of information show the most dependence (influence) and the variable of the commodity factor is the only It is a variable that has a two-way relationship with other variables. Manuscript profile -
Open Access Article
23 - Feature Selection And Clustering By Multi-objective Optimization
Seyedeh Mohtaram Daryabari Farhad Ramezani -
Open Access Article
24 - FFS: A F-DBSCAN Clustering- Based Feature Selection For Classification Data
Nasim Eshaghi Ali Aghagolzadeh -
Open Access Article
25 - A New Hybrid Model of K-Means and Naïve Bayes Algorithms for Feature Selection in Text Documents Categorization
Ali Allahverdipour Farhad Soleimanian Gharehchopogh -
Open Access Article
26 - An Improved Flower Pollination Algorithm with AdaBoost Algorithm for Feature Selection in Text Documents Classification
Hiwa Majidpour Farhad Soleimanian Gharehchopogh -
Open Access Article
27 - A Parallel Genetic Algorithm Based Method for Feature Subset Selection in Intrusion Detection Systems
Iran Shokripoor Bahman Bigloo -
Open Access Article
28 - A Novel Approach to Feature Selection Using PageRank algorithm for Web Page Classification
Farhad Rezvani Farhad Soleimanian Gharehchopogh -
Open Access Article
29 - NSE: An effective model for investigating the role of pre-processing using ensembles in sentiment classification
Razieh Asgarnezhad Amirhassan Monadjemi -
Open Access Article
30 - Improving Short-Term Wind Power Prediction with Neural Network and ICA Algorithm and Input Feature Selection
Elham Imaie Abdolreza Sheikholeslami Roya Ahmadi Ahangar -
Open Access Article
31 - A Low Complexity ANFIS Approach for Premature Ventricular Contraction Detection Based on Backward Elimination
Zahra Sadeghi Hamid Jazayeriy Soheil Fateri -
Open Access Article
32 - A New Approach in Persian Handwritten Letters Recognition Using Error Correcting Output Coding
Maziar Kazemi Muhammad Yousefnezhad Saber Nourian -
Open Access Article
33 - A New Multi-Stage Feature Selection and Classification Approach: Bank Customer Credit Risk Scoring
Farshid Abdi -
Open Access Article
34 - Motor Signal Intelligent Processing in Huntington Disease Diagnosis
Mohammad Karimi Moridani Soroor Behbahani Sepeideh Asadikia -
Open Access Article
35 - Determining the effective features in classification of heart sounds using trained intelligent network and genetic algorithm
mahsa semyari fardad farokhi -
Open Access Article
36 - Bionic Wavelet Transform Entropy in Speaker-Independent and Context-Independent Emotional State Detection from Speech Signal
Mina Kadkhodaei Elyaderani Hamid Mahmoodian -
Open Access Article
37 - Determining Effective Features for Face Detection Using a Hybrid Feature Approach
Sepideh Araban Fardad Farokhi Kaveh Kangarloo -
Open Access Article
38 - Effective Feature Selection for Pre-Cancerous Cervix Lesions Using Artificial Neural Networks
Farnaz Rouhbakhsh Fardad Farokhi Kaveh Kangarloo -
Open Access Article
39 - Neuro-Fuzzy Based Algorithm for Online Dynamic Voltage Stability Status Prediction Using Wide-Area Phasor Measurements
Ahmad Ahmadi Yousef Alinezhad Beromi -
Open Access Article
40 - Identification of effective indicators on predicting trends of total index of Tehran Stock Exchange using feature selection and classification algorithms
Mohammad Javad Sheikhzadeh Sajjad RahmanyBecause of the numerous environmental, industrial, micro, and macro elements that influence the index and stock price trend, it is undeniably difficult to predict changes in the index and stock price trend. Although the aforementioned factors are difficult or impossible MoreBecause of the numerous environmental, industrial, micro, and macro elements that influence the index and stock price trend, it is undeniably difficult to predict changes in the index and stock price trend. Although the aforementioned factors are difficult or impossible to measure in some cases, micro factors such as price history and trade volume are simply measurable and available. The goal of this research is to use feature selection and classification algorithms to find the most effective features and indicators for predicting the total index and total weighted index. Then we will examine the proposed model in the daily and weekly time period from January 2020 to May 2022. The results show that it is possible to predict the trend of changes with high accuracy using a limited number of indicators, and that the prediction accuracy is much higher in the weekly time interval than in the daily time interval. Manuscript profile -
Open Access Article
41 - Predicting the daily index of the Tehran Stock Exchange using the selection of appropriate features for the Long Short-Term Memory neural network (LSTM)
Somayeh Mohebi Mohammad Esmaeil Fadaeinejad mohammad osoolian Mohammad reza HamidizadehThe stock market index is one of the effective features in investment because it can well reflect the health status and macro change trend of a country’s economic development. Various features affect the stock index. The various combinations of these features crea MoreThe stock market index is one of the effective features in investment because it can well reflect the health status and macro change trend of a country’s economic development. Various features affect the stock index. The various combinations of these features create a wide state space. Hence, it is impractical to provide a data set containing all these combinations to train the stock index prediction model. in this research, an attempt has been made, after collecting a significant number of effective features on the index, to provide a method for selecting appropriate features for the stock index prediction model with aim of increasing prediction accuracy. For this purpose, the mRMR algorithm is used as the basic algorithm. Also, to select the appropriate model, a number of the most applicable artificial intelligence models for predicting the stock index were compared and according to the results, the LSTM network was selected to predict the stock index. The results of this study show that using the LSTM network and the proposed method in selecting features, with 8 selected features, high accuracy can be achieved in the daily prediction of the Tehran Stock Exchange Index. So that MPE is calculated to be about 2.66, Manuscript profile -
Open Access Article
42 - Portfolio Formation Using Diagonal Quadratic Discriminant Analysis and Weighting Based on Posterior Probability
Saeid Fallahpour H. Pirayesh ShirazinejadStock return forecasting is one of the most important question for investing in Stock markets. Because of the effects of policy, economic, etc., we need moderns and intelligent models to forecast the returns. The main idea in this research is classifying the stocks int MoreStock return forecasting is one of the most important question for investing in Stock markets. Because of the effects of policy, economic, etc., we need moderns and intelligent models to forecast the returns. The main idea in this research is classifying the stocks into high and low return groups, for this purpose support vector machine (SVM) was used. To elect the best variables for models we used sequential feature selection and in order to evaluate the accuracy of SVM we do the same forecasting with diagonal quadratic discriminant analysis (DQDA). By using paired t-test, we conclude that models have no significant difference. Equal weighted portfolios were created for each models with and without feature selection also, we used posterior probability to weight the portfolio of DQDA with feature selection. The returns were calculated for each portfolio during the years 1388-1391. The simulating results are satisfying and all portfolios’ returns are better than market portfolio. Manuscript profile -
Open Access Article
43 - Comparing Different Feature Selection Methods in Financial Distress Prediction of the Firms Listed in Tehran Stock Exchange
Mohammad Namazi Mostafa Kazemnezhad M. Mahdi NematollahiResearch in financial distress and bankruptcy emphasize the design of more sophisticated classifiers, and less feature (variables) selection and their appropriate methods. In this regard, the purpose of this study is to compare performance of different feature selection MoreResearch in financial distress and bankruptcy emphasize the design of more sophisticated classifiers, and less feature (variables) selection and their appropriate methods. In this regard, the purpose of this study is to compare performance of different feature selection methods in financial distress prediction of the companies listed on Tehran Stock Exchange (TSE). In this regard, we investigated and compared five feature selection methods, including t-test, stepwise regression, factor analysis, relief, wrapper subset selection and RFE-SVM feature selection. To obtain comparable experimental results (reliable comparison), three different classifiers (including neural networks, support vector machine and AdaBoost) were used in this study. In overall, the experimental results confirmed the usefulness of variable selection methods and significant difference among level (amount) of different methods performance. In other words, the application of the feature selection methods increases the mean of accuracy, and reduces the occurrence of type I and type II errors. Furthermore, the results indicated that wrapper subset selection method outperforms the other feature selection methods. Manuscript profile -
Open Access Article
44 - Smart car system: automobile driver's stress recognition with artificial neural networks
Mahtab Vaezi Mehdi Nasri Farhad Azimifar -
Open Access Article
45 - A hybrid bankruptcy prediction model based on GMDH-type neural network and genetic algorithm for Tehran Stock Exchange companies
hosain vazifehdost tayebeh zangenehThis paper proposes a Soft Computing model for effective bankruptcy prediction, based on the integration of Group Method of Data Handling (GMDH) neural network and genetic algorithm which is called here as GA-GMDH. Genetic algorithm (GA) designs the whole architec MoreThis paper proposes a Soft Computing model for effective bankruptcy prediction, based on the integration of Group Method of Data Handling (GMDH) neural network and genetic algorithm which is called here as GA-GMDH. Genetic algorithm (GA) designs the whole architecture of the GMDH network and optimizes its topology. In order to demonstrate the effectiveness of our proposed GA-GMDH model, its performance was compared with performance of the commonly used statistical techniques of logistic regression (LR) and a relatively new artificial intelligent technique of Adaptive Neuro-Fuzzy Inference System (ANFIS). Performance of the designed prediction models depends on the utilized variable selection technique. Therefore, we constructed 12 prediction models through combining the four filtering feature selection methods and the three prediction models. The four feature selection methods of independent samples T-test, correlation matrix (CM), stepwise multiple discriminant analysis (SDA) and principal component analysis (PCA)are combined with prediction models to generate four optimal feature subsets. Empirical data were collected one year prior to failure from Tehran Stock Exchange (TSE) during 1997-2008. For robust assessing of prediction models’ performance, we applied Type-I and Type-II errors, and area under the receiver operative characteristics curve (AUC) measures. Experimental results indicate that our proposed GA-GMDH model has high ability in bankruptcy prediction problem and significantly outperforms ANFIS and LR models in all combinations with four feature selection methods. Meanwhile, the CM method has the best ability in selecting predictive variables in comparison with other feature selection methods. Therefore, CM-GA-GMDH model is determined as the best constructed model for bankruptcy prediction using our particular dataset from TSE. Manuscript profile -
Open Access Article
46 - Optimal Feature Selection for Data Classification and Clustering: Techniques and Guidelines
Farhad Rad Ali Asghar Nadri Hamid Parvin