Abstract:
Time series models have been utilized to make reasonably accurate predictions in the
areas of stock price movements, academic enrollments, weather and many more. For
promoting the forecasting performance of fuzzy time-series models, Tai-Liang Chen,
Ching-Hsue Cheng and Hia-Jong Teoh had proposed a new model, which incorporates
the concept of the Fibonacci sequence, the framework of Song and Chissom's model
and the weighted method of Yu's model. Their findings were shown in the journal
entitled "Fuzzy time-series based on Fibonacci sequence for stock price forecasting
(2007)". We noticed that length of intervals somehow affects the performance of
fuzzy time series as proposed by Huamg (2000) who argued that different lengths of
intervals lead to different forecasting results and forecasting errors. Consequently, it
affects the performance of the model proposed by Chen, T.L., Cheng and Teoh (2007).
Therefore, we employ the frequency-density-based partitioning into their model in
order to compare it with its original randomly chosen length of intervals partitioning.
This paper employs a 2-year weekly period of Kuala Lumpur Composite Index (KLCI)
stock index data as experimental datasets. Through comparison of the forecasting
performances of our model with their model, we noticed that our model has smaller
forecasting error. Hence, conclude that our model is an improved model of the model
proposed by Chen, T.L., Cheng and Teoh (2007) previously.
