Abstract
Time-Series models are used for predictions in whether forecasting, academic enrollments,
rice production etc. The concept of fuzzy time series is introduced by Song & Chissom in 1993. Over
the past 24 years, many fuzzy time series methods have been proposed for rice production forecasting,
but the forecasting accuracy rate of the existing methods is not good enough. These methods have used
actual production, difference of production or percentage change in production as the universe of
discourse. And frequency density based partitioning or ratio based partitioning for partition of
discourse. This paper, proposed a method based on fuzzy time series, which gives the more accurate
result than the existing methods. The proposed method used the actual production as the universe of
discourse and mean based partitioning as partition of discourse. To illustrate the forecasting process,
the historical data of rice production of University of Agriculture and Technology, India is used.