Neural Network

The main purpose of this paper is providing a model to calculate the credit risk of Melli bank clients and implement it at Mellat Bank. Therefore, the present study uses a multi-layered neural network method. The statistical population of this research is all real and legal clients of Melli and Mellat banks. Sampling method used in this research is a simple random sampling method. Friedman test was used to calculate the required number of samples in a random sampling method from Cochran formula (1977) and Friedman test was used to rank the factors affecting the credit risk. Friedman test was also performed using data from a completed questionnaire of active experts at the Melli Bank. Based on the results obtained from Friedman test, five important factors in the credit risk of real clients of the Melli Bank of Iran, type of occupation, guarantee value, loan amount, having return checks, the balance average, and the value of the guarantee, the amount of the loan, the average of the balance, having returned checks and deferred loans are the most important factors affecting the credit risk of legal clients, which have been used as inputs in the neural network model. The results of credit risk prediction using the neural network showed that the designed model has a high ability to predict the credit risk of real and legal clients of the Melli bank, while it did not have this ability for the Mellat bank.

In this paper, a model based on GMDH Type Neural Network, is used to predict gas price in the spot market while using oil spot market price, gas spot market price, gas future market price, oil future market price and average temperature of the weather. The results suggest that GMDH Neural Network model, according to the Root Mean Squared Error (RMSE) and Direction statistics (Dstat) statistics are more effective than OLS method. Also, first lag of gas price in the future market is the most efficient variable in predicting gas price in spot market.

In recent years, authors have focused on modeling and forecasting volatility in financial series it is crucial for the characterization of markets, portfolio optimization and asset valuation. One of the most used methods to forecast market volatility is the linear regression. Nonetheless, the errors in prediction using this approach are often quite high. Hence, continued research is conducted to improve forecasting models employing a variety of techniques. In this paper, we extend the field of expert systems, forecasting, and model by applying an Artificial Neural Network. ANN model is applied to forecast market volatility. The results show an overall improvement in forecasting using the neural network as compared to linear regression method.

Successful trades in financial markets have to be conducted close to the key recurrent points. Researchers have recently developed diverse systems to help the identification of these points. Technical analysis is one of the most valid and all-purpose kinds of these systems. With its numerous rules, the technical analysis endeavors to create well-timed and correct signals so that these points are identified. However, one of the drawbacks of this system is its overdependence on human analysis and knowledge in selecting and applying these rules. Employing the three tools of genetic algorithm, fuzzy logic, and neural network, this study attempts to develop an intelligent trading system based on the recognized rules of the technical analysis. Indeed, the genetic algorithm will assist with the optimization of technical rules owing to computing complexities. The fuzzy inference will also help the recognition of the total current condition in the market. It is because a set of rules will be selected based on the market kind (trending or non-trending). Finally, the signal developed by every rule will be translated into a single result (buy, sell, or hold). The obtained results reveal that there is a statistically meaningful difference between a stock's buy and hold and the trading system proposed by this research. In other words, our proposed system displays an extremely higher profitability potential.

Today, social networks are fast and dynamic communication intermediaries that are a vital business tool. This study aims at examining the views of those involved with Facebook stocks so that we can summarize their views to predict the general behavior of this stock and collectively consider possible Facebook stock price movements, and create a more accurate pattern compared to previous patterns. In this study, we have analyzed two statistical samples, the first being a large dataset containing a variety of tweets with an emotional tag. That is, it needed a set that had already been extracted from each individual tweet by a trusted human or machine. Consequently, we have collected posts on Facebook in an eighty-day period. In this study, we used a tagged dataset using Python's programming language and vector-to-word algorithm. The research results show that, we need stock change information, machine learning and sentiment analysis, and on paper we conclude that positive news about a company excites people to have positive opinions about it which in turn results in people encouraging each other to buy and hold stocks. Meanwhile, the opposite trend is also true, but everything will not always be easy and clear, and it is in areas of high complexity and mental uncertainty that the art of using the three elements mentioned above is evident.

The conjecture of stock exchange is the demonstration of attempting to decide the forecast estimation of a particular sector or the market, or the market as a whole. Every stock every investor needs to foresee the future evaluation of stocks, so a predicted forecast of a stock’s future cost could return enormous benefit. To increase the accuracy of the Conjecture of stock Exchange with daily changes in the market value is a bottleneck task. The existing stock market prediction focused on forecasting the regular stock market by using various machine learning algorithms and in-depth methodologies. The proposed work we have implemented describes the new NN model with the help of different learning techniques like hyperparameter tuning which includes batch normalization and fitting it with the help of random-search-cv. The prediction of the Stock exchange is an active area for research and completion in Numerai. The Numerai is the most robust data science competition for stock market prediction. Numerai provides weekly new datasets to mold the most exceptional prediction model. The dataset has 310 features, and the entries are more than 100000 per week. Our proposed new neural network model gives accuracy is closely 86%. The critical point, it isn’t easy with our proposed model with existing models because we are training and testing the proposed model with a new unlabeled dataset every week. Our ultimate aim for participating in Numerai competition is to suggest a neural network methodology to forecast the stock exchange independent of datasets with reasonable accuracy.

The tremendous advances in artificial intelligence over the past decade have led to their increasing use in financial markets. In recent years a large number of investment companies and hedge funds have been implementing algorithmic and automated trading on their trading. The speed of decision-making and execution is the most important factor in the success of institutional and individual investors in capital markets. Algorithmic trading using machine learning methods has been able to improve the performance of investors by finding investment opportunities as well as time entry and exit of trading. The purpose of this study is to achieve a better portfolio performance by designing an intelligent and fully automated trading system that investors with the support of this system, in addition to finding the best opportunities in the market, can allocate resources optimally. The present study consists of four separate steps. Respectively, tuning the parameters of technical indicators, detecting the current market regime (trending or non-trending), issuing a definite signal (buy, sell or hold) from the indicators’ signals and finally portfolio rebalancing. These 4 steps respectively are performed using genetic algorithm, fuzzy logic, artificial neural network and conventional portfolio optimization model. The results show the complete superiority of the proposed model in achieving higher returns and less risk compared to the performance of the TEDPIX and other mutual funds in the same period.

In this paper, a fuzzy ontology for Semantic Sensor Networks (SSN) is proposed for smart homes in two phases. In the first phase, using the WordNet ontology, the location and type of an object is identified with the aid of a graphical interface. This object and its synonyms are added to the list of the known objects set. Succeeding, the relation of the object with other groups is assessed based on a similarity measure in addition to using the fuzzy ontology. In the second phase, sensors with erroneous information are identified and pruned by finding a relationship between some specific factors. To this end, temperature, moisture and light are considered and the Adaptive Neuro-Fuzzy Inference System (ANFIS) is incorporated. The proposed method is implemented using some parts of the Wikipedia database and the WordNet dictionary. The first phase of the proposed method is tested with several sample requests and the system shows favorable results on finding the original group (and other related groups) of the request. For training the neural network in the second phase, the Intel lab Dataset is used. Results of this phase show that the neural network can predict the temperature and moisture factors with low error, while the light factor has more error in prediction

The issue of asset pricing in the market is one of the most important and old issues in the financial world. Factor pricing models seek to be able to determine a significant relationship between return on assets based on the risk parameters of that asset. A wide range of factors can be found in the literature that can be an element for measuring the risk of an asset, but the big question is which of these models will work better. The factors studied in this research include factors that cover market risk, valuation risk, psychological (technical) market risk, profit quality risk, profitability, investment, etc. In this study, we have tried to Using machine learning techniques and optimization tools is a way to derive adaptive-robust nonlinear models that can reduce the risk of model error as much as possible. In this research, two models have been developed. In the first model, using the feature extraction technique and optimization of models based on neural network, a non-linear and adaptable model has been developed for each asset. In the second approach, a portfolio of improved neural network-based models is used in the first stage, which can be used to minimize the risk of model error and achieve a model that is resistant to different market conditions. Finally, it can be seen that the development of these models can significantly improve the risk of error and average error of the model compared to traditional CAPM approaches and the Fama and French three-factor model.

Emotion recognition from speech has noticeable applications within the speech-processing systems. The goal of this paper is to permit a totally natural interaction among human and system. In this paper, an attempt is made to design and implement a system to determine and detect emotions of anger and happiness in the Persian speech signals. Research on recognizing some emotions has been done in most languages, but due to the difficulty of creating a speech database, so far little research has been done to identify emotions in Persian speech. In this article, because of the dearth of a suitable database in Persian to detect feelings, before everything, a database for moods of happiness and anger and neutral (with no emotion) in Persian, including 720 sentences was set up. Then the frequency features of speech signals obtained from Fourier transform such as maximum, minimum, median and mean as well as LPC coefficients were extracted. Then, the MLP neural network was used to detect emotions of happiness and anger. Results show that our algorithm performs 87.74% accurately.

The electronic industry suffers a rapid changing and highly rival environment. Thus, firms have an essential need to strive for acquiring the competitive advantage. Strategy Organizational Agility (SOA) is a tool which enables to assist firms to attain the competitive advantage. Therefore, this study benchmarks the core competencies from a case study within the supply chain network and establishes a set of attributes for augmenting SOA. A novel multi-criteria decision-making structure is proposed to deal with the complex interrelationships among the aspects and attributes. Radial basis function (RBF) neural network can use linear learning algorithm to complete the work formerly handled by nonlinear learning algorithm, and maintain the high precision of the nonlinear algorithm. However, the results of RBF would be slightly unsatisfactory when dealing with small sample which has higher feature dimension and fewer numbers. Higher feature dimension will influence the design of neural network, and fewer numbers of samples will cause network training incomplete or over-fitted, both of which restrict the recognition precision of the neural network. The competence and accountability indicators can the continuously increasing level of agility to be effective. According to the analysis chart production and product design performance indicators alone cannot level a considerable amount of agility to change. But reducing the level of the index level of agility is reduced. Flexibility indices speed and agility level changes can also affect the organization. But with increased levels of these two indicators increased agility rate change indicators will be more flexibility. The results showed that customer knowledge management impact on organizational agility and organizational effectiveness and customer knowledge management through organizational agility has significant positive impact on organizational effectiveness. Finally, some practical suggestions, future research suggestions and research limitations are presented.

Annually, researchers in various scientific fields publish their research results as technical reports or articles in proceedings or journals. The collocation of this type of data is used by search engines and digital libraries to search and access research publications, which usually retrieve related articles based on the query keywords instead of the article’s subjects. Consequently, accurate classification of scientific articles can increase the quality of users’ searches when seeking a scientific document in databases. The primary purpose of this paper is to provide a classification model to determine the scope of scientific articles. To this end, we proposed a model which uses the enriched contextualized knowledge of Persian articles through distributional semantics. Accordingly, identifying the specific field of each document and defining its domain by prominent enriched knowledge enhances the accuracy of scientific articles’ classification. To reach the goal, we enriched the contextualized embedding models, either ParsBERT or XLM-RoBERTa, with the latent topics to train a multilayer perceptron model. According to the experimental results, overall performance of the ParsBERT-NMF-1HT was 72.37% (macro) and 75.21% (micro) according to F-measure, with a statistical significance compared to the baseline (p<0.05).

Contemporary machine learning models, like deep neural networks, require substantial labeled datasets for proper training. However, in areas such as natural language processing, a shortage of labeled data can lead to overfitting. To address this challenge, data augmentation, which involves transforming data points to maintain class labels and provide additional valuable information, has become an effective strategy. In this paper, a deep reinforcement learning-based text augmentation method for sentiment analysis was introduced, combining reinforcement learning with deep learning. The technique uses Deep Q-Network (DQN) as the reinforcement learning method to search for an efficient augmentation strategy, employing four text augmentation transformations: random deletion, synonym replacement, random swapping, and random insertion. Additionally, various deep learning networks, including CNN, Bi-LSTM, Transformer, BERT, and XLNet, were evaluated for the training phase. Experimental findings show that the proposed technique can achieve an accuracy of 65.1% with only 20% of the dataset and 69.3% with 40% of the dataset. Furthermore, with just 10% of the dataset, the method yields an F1-score of 62.1%, rising to 69.1% with 40% of the dataset, outperforming previous approaches. Evaluation on the SemEval dataset demonstrates that reinforcement learning can efficiently augment text datasets for improved sentiment analysis results.