Teimour Mohammadi; Atefeh Taklif; Sahel Zamani
Abstract
In this article, we introduce a model for forecasting the daily gas prices by the use of wavelet transform and neural networks. In this hybrid model, the discrete Daubechies wavelet transform is applied to decompose the gas prices series into approximation series and details series (DS). The new series ...
Read More
In this article, we introduce a model for forecasting the daily gas prices by the use of wavelet transform and neural networks. In this hybrid model, the discrete Daubechies wavelet transform is applied to decompose the gas prices series into approximation series and details series (DS). The new series are used as inputs to the ANN model to forecast Henry Hub natural gas prices. The relative performance of the hybrid model and neural network model shows that WANN model provides more accurate naturel gas price forecast compared to the individual ANN model. Diebold-Mariano test confirms this result.
Farzad Eskandari; Ghazaleh Baghbani
Abstract
Banks, on the one hand are involved with the challenge of inadequate cash to meet the customers’ needs and on the other hand, are reluctant to increase the costs resulting from the cash excess transfer. As a result, estimating the cash requirements of the bank's branches, according to their daily ...
Read More
Banks, on the one hand are involved with the challenge of inadequate cash to meet the customers’ needs and on the other hand, are reluctant to increase the costs resulting from the cash excess transfer. As a result, estimating the cash requirements of the bank's branches, according to their daily operations, which is considered as a multivariable system, is one of the most important issues in banking. In this regard, employing data mining, especially clustering methods and neural networks can help to increase the accuracy of estimating the cash required in branches. In this regard, Neural networks are considered significant in terms of flexibility, nonlinearity, greater tolerance to noise and independence from the basic assumptions about the input data.
In the present paper, 20 branches of Tejarat bank have been categorized in similar clusters, during the period 21/04/2014 and 22/09/2014, according to factors such as branch grade, the type of branches in terms of deposit or facility, the number of ATMs, stand-by branches. Then, considering the clustering results and the variables related to the cash of branches such as week days, payment/ deposit subsidy/ deposit interest days, holidays and official events, as well as the amount of cash used in ATMs, the suitable structure for the neural network has been identified to estimate the required cash via the error criteria and the required cash is accordingly estimated for different clusters. The results show that the neural network, considering the clustering results, can estimate the required cash of branches in different clusters with good performance with a mean absolute error of 5%.
Ali Mohamad Ahmadi; Mahdi Zolfaghari; Aidin Ghafar Nejad Mehrabani
Volume 13, Issue 41 , February 2010, , Pages 107-121
Abstract
Electricity demand is growing very fast in Iran and it is important to forecast its future demand and its monthly variation accurately. Artificial Neural Network (ANN) is a powerful tool for nonlinear models for forecasting and it was used to estimate monthly electricity demand in this study. In this ...
Read More
Electricity demand is growing very fast in Iran and it is important to forecast its future demand and its monthly variation accurately. Artificial Neural Network (ANN) is a powerful tool for nonlinear models for forecasting and it was used to estimate monthly electricity demand in this study. In this paper, we compared the Non-linear ANN model with ARIMA linear model to estimate monthly electricity demand for a priod of 3 years. Using MSE, RMSE, NMSE, MHE, MAPE and R2 indicatorss, our results show that ANN forecasting model is superior to ARIMA in terms of less error coefficient and high explanatory ability.
Hamid Nilsaz; Abdolrahman Rasekh; Alireza Osareh; Hasanali Sinae
Volume 9, Issue 32 , October 2007, , Pages 85-109
Abstract
Traditional methods of deciding whether to grant credit to a particular individual use human judgment of the risk of default based on experience of previous decisions. However, economic pressures resulting from increased demand for credit, allied with greater commercial competition and the emergence ...
Read More
Traditional methods of deciding whether to grant credit to a particular individual use human judgment of the risk of default based on experience of previous decisions. However, economic pressures resulting from increased demand for credit, allied with greater commercial competition and the emergence of new computer technology have led to development of sophisticated statistical models to aid the credit granting decision making process. Credit scoring is the name used to describe this process of determining how likely applicants are to default with their repayments. Credit scoring has some obvious benefits that have led to its increasing use in loan evaluation. For example, it is quicker, cheaper and more objective than judgmental method. A wide range of statistical methods such as discriminant analysis, logistic regression, and neural networks have been applied for credit scoring. In this paper, we design a neural network credit scoring system for classifying the applicants of personal loans in bank and compare the performance of this model with discriminant analysis and logistic regression models. The results of this investigation show that the neural network model is more accurate and more flexible than discriminant analysis and logistic regression.