決策樹結合複迴歸模型預測氣溫與雨量

Abstract

本論文發展一以決策樹為基礎的氣溫與雨量之預測模型。由於傳統決策樹都是以區間化輸出為主，因此該預測模型將線性複迴歸模型整合至決策樹以達成數值化輸出。本論文利用該預測模型來預測領前1至7天的氣溫與雨量，並針對預測值給予一定信賴水準的信賴區間。為了說明此預測模型的效能，該預測模型與其他時間序列的預測方法進行比較，其中時間序列的預測方法包括自迴歸、移動平均法、自迴歸差分整合移動平均法。

Based on decision tree, the purpose of this thesis is to develop a forecast model for temperature and rainfall. Because the traditional decision tree generates interval output, the forecast model integrates the multiple linear regression model into the decision tree in order to achieve the goal of numeric output. In this thesis, the seven days ahead temperature and rainfall are predicted by using the forecast model, and their confidence intervals are given at a confidence level. In order to demonstrate the effectiveness of the forecast model, we compare the forecast model with some different time series methods, such as autoregressive (AR), moving average(MA), autoregressive integrated moving average (ARIMA).