IJCSNS International Journal of Computer Science and Network Security, VOL.8 No.9, September 2008 382 Manuscript received September 5, 2008. Manuscript revised September 20, 2008. A Comparative Study of Neural-Network & Fuzzy Time Series Forecasting Techniques – Case Study: Wheat Production Forecasting Adesh Kumar Pandey*, A.K Sinha**, and V.K Srivastava*** *Krishna Institute of Engineering & Technology, Associate Professor, Ghaziabad, India, **Krishna Institute of Engineering & Technology, Professor, Ghaziabad, India, **Krishna Institute of Engineering & Technology, Professor, Ghaziabad, India, Summery Various forecasting methods have been developed on the basis of fuzzy time series data, but accuracy has been matter of concern in these forecasts. As in fuzzy time series methods forecasted values depend to some degree on our interpretation of the output of the forecasting model thus different interpretation may lead to different results, this makes the process quite subjective. An objective method, based on artificial neural network of forecasting is proposed .The proposed method is compared with various fuzzy time series forecasting methods. Key words Agriculture Production, Forecasting, Fuzzy time series, Neural Network 1. Introduction Forecasting the behavior of complex system has been a broad application domain for neural networks. In particular, such as electric load forecasting [1], [2], economic forecasting [3], forecasting natural physical phenomena [4] , river flow forecasting [5] and forecasting student admission in colleges[18] have been widely studied. Other than neural network based forecasting, Fuzzy time series forecasting emerged as a noble approach for predicting the future values in a situation where neither a trend is viewed nor a pattern in variations of time series are visualized and moreover the information (data) are imprecise and vague. Song and Chissom [6] successfully employed the concept of fuzzy sets having linguistic variables presented by Zadeh [11, 12] and the application of fuzzy logic to approximate reasoning by Mamdani [13] to develop the foundation of fuzzy time series forecasting. Song and Chissom [13, 14] implemented his developed time invariant and time variant models on the historical time series data of student enrollments of university of Alabama. Chen [14] presented a simplified time invariant method for time series forecasting by using the arithmetic operations in place of max-min composition operation used by Song and Chissom [7]. Further, Chen [14] applied the high order fuzzy time series model for forecasting the enrolments and found some points of ambiguity to the trends in forecast and suggested to use high order fuzzy logical relationship group to deal with ambiguity. S.R. Singh [10] presented an improved and versatile method for fuzzy time series forecasting using a difference parameter as fuzzy relation for forecasting. Rajesh Joshi [17] used a fuzzy time series model for agricultural production forecasting comprises of the development and implementation of fuzzy series model using metrological parameters as indicators for forecasting. In this paper to achieve the objectivity over the subjectivity of fuzzy time series based methods a neural network based methods has been proposed. The proposed method has been implemented on the historical data and influencing parameters (temperature, sunshine and rainfall) of crop (wheat) production of Pant Nagar farm, G.B. Pant University of agriculture and technology, Pant Nagar (India) [17] Agriculture production system is one of the real life problems falling in the category having uncertainty in known and some unknown parameters, hence become a natural choice for implementation of fuzzy time series forecasting models in its production system. The uncertainty lies in the crop production due to some uncontrolled parameters of, which ‘weather’, ‘agro meteorological’ variables are key contents. Further, the crop production being dealt with the field data, precision of data is always a matter of concern. Past experience shows that the crop production system may observe the large variation in production data as the system is effected by many uncertain production parameters and uncertain occurrence of natural calamities. The proposed method produces better result than above discussed fuzzy time series based methods. 2. Artificial Neural Networks and Fuzzy Time Series In view of making our study self explanatory, same basic definition and properties of fuzzy time series and neural network g found in [1 – 18] are presented as: