Neural Networks

The issue of credit risk and deferred bank claims is one of the sensitive issues of banking industry, which can be considered as the main cause of bank failures. In recent years, the economic slowdown accompanied by inflation in Iran has led to an increase in deferred bank claims that could put the country's banking system in serious trouble. Accordingly, the current paper presents a prediction model for credit risk of real customers of Qavamin Bank Branch in Shiraz, using a combined approach of logistic regression and neural network. Therefore, the necessary examinations were carried out on a sample of 351 individuals from the real customers of the bank in the period 2011-2012. According to the information available, 17 variables were extracted including financial and non-financial variables for classifying customers into well-balanced s and ill-balanced s. Among the variables, five effective variables on credit risk were selected using the parent forward stepwise selection technique, which was used to train neural networks with three neurons in the hidden layer. the optimum cutting point was selected based on the performance curve of the system and the results of the neural network output on the test data show that the accuracy of the combined model in the classifier of well-balanced customers is .89 and in the category of ill-balanced customers is .83 that is better than the results of logistic regression and in general, it is possible to estimate the accuracy of prediction.

Stock market forecasting is a challenging task for investors and researchers in the financial market due to highly noisy, nonparametric, volatile, complex, non-linear, dynamic and chaotic nature of stock price time series. With the development of computationally intelligent method, it is possible to predict stock price time series more accurately. Artificial neural networks (ANNs) are one of the most promising biologically inspired techniques. ANNs have been widely used to make predictions in various research. The performance of ANNs is very dependent on the learning technique utilized to train the weight and bias vectors. The proposed study aims to predict daily Tehran Exchange Dividend Price Index (TEDPIX) via the hybrid multilayer perceptron (MLP) neural networks and metaheuristic algorithms which consist of genetic algorithm (GA), particle swarm optimization (PSO), black hole (BH), grasshopper optimization algorithm (GOA) and grey wolf optimization (GWO). We have extracted 18 technical indicators based on the daily TEDPIX as input parameters. Therefore, the experimental result shows that grey wolf optimization has superior performance to train MLPs for predicting the stock market in metaheuristic-based.

Community Question Answering has a crucial role in almost all societies nowadays. It is important for the owners of a community to be able to make it better and more reliable. One way to achieve this, is to find the users who have more knowledge, expertise, experience and skill and can well share their knowledge with others (which we call experts and aim to encourage them to be more active in the website). One method to use is to identify expert users, and whenever a new question is asked, we suggest this question to them to check and answer if its in their area of expertise. One way to encourage users to post replies, is to use gameplay techniques such as assigning points and badges to users. But as we will discuss, this method does not always detect expert users well, because some users will try to have small and insignificant but numerous activities that will make them gain a lot of points, however they are not experts. In this study, we examine the methods by which experts in a question-and-answer system can be found, and try to evaluate and compare these methods, use their ideas and positive points, and add our own new ideas to a new way of finding them. We used some ideas such as profile making for users, categorize users’ expertise, A-Priori algorithm and showed that neural networks method results the best for the purpose of expert detection.

Nowadays, the tourism industry has become one of the most important sectors in the world economy. Due to the perishability of this industry, accurate forecasting of the demand is very important for tourism planning and resource allocation. Studies show that due to the diversity and complexity of the factors affecting tourism demand, the combination of different approaches may increase the forecasting accuracy. The aim of this paper is to forecast the tourism demand of Alisadr cave. For this purpose, a method based on artificial neural networks is presented, in which the results of linear and non-linear methods and short-term and long-term forecasts are combined. This method is applied to a dataset of Alisadr cave tourists. The evaluation results show that in most cases, the proposed combined method can predict the tourism demand with higher accuracy than the monthly and seasonal methods based on neural networks and random forest models. The predictive models obtained from this study can enhance customer service and improve the interaction between users and tourist ticketing web applications and online reservation programs.

Heart and circulatory system diseases are often referred to as cardiovascular disease (CVD). The health and efficiency of the heart are crucial to human survival. CVD has become a primary cause of demise in recent years. According to data provided by the World-Health-Organization (WHO), CVD were conscientious for the deaths of 18.6M people in 2017. Biomedical care, healthcare, and disease prediction are just few of the fields making use of cutting-edge skills like machine learning (ML) and deep learning (DL). Utilizing the CVD dataset from the UCI Machine-Repository, this article aims to improve the accuracy of cardiac disease diagnosis. Improved precision and sensitivity in diagnosing heart disease by the use of an optimization algorithm is possible. Optimization is the process of evaluating a number of potential answers to a problem and selecting the best one. Support-Machine-Vector (SVM), K-Nearest-Neighbor (KNN), Naïve-Bayes (NB), Artificial-Neural-Network (ANN), Random-Forest (RF), and Gradient-Descent-Optimization (GDO) are just some of the ML strategies that have been utilized. Predicting Cardiovascular Disease with Intelligence, the best results may be obtained from the set of considered classification techniques, and this is where the GDO approach comes in. It has been evaluated and found to have an accuracy of 99.62 percent. The sensitivity and specificity were likewise measured at 99.65% and 98.54%, respectively. According to the findings, the proposed unique optimized algorithm has the potential to serve as a useful healthcare examination system for the timely prediction of CVD and for the study of such conditions.