Research on Constructing Product Sales Forecasting Mode by Using Time Series Algorithms – Taking HiNet Virtual Point Card as an Example

• Main Authors: TSAI, YUEH-CHIH, 蔡岳志
• Other Authors: LIOU, JIA-HUA
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/z88ksg

碩士 === 醒吾科技大學 === 資訊科技應用系 === 105 === In recent years, along with the globalization trends in market development, businesses are facing more and more intense market competition. For a business to stand out from the market, a key factor of competitive companies is the ability to spend the least amount of time correctly predicting future market trends. The ways in which businesses have conducted effective analysis of their product sales to further make effective decisions in line with market trends, have become a key for company survival. Through systematic analysis, general marketing information can provide insightful business intelligence for decision-making, which is an urgently needed business skill. The process of making a sales forecast means estimating amount of sales of a specific product, or group of products, within sales a specified period in the future. Sales forecasting is important as it is the basis for companies to evaluate their future business direction and policy decisions. This study collected the information from electronic invoice receipts, and after screening obtained the information from "HiNet Point Card" historical sales data, as the analysis target of this study. First, the historical data was divided into two parts: the "training set" and the "verification data." The "training set" was added into the Weka data investigation tool platform, and then time series algorithms were used to generate forecast data. Afterwards, the forecast data was then compared with the "verification data" data by comparative analysis using the sample T test. The results indicated that the Linear Regression algorithm has the best predictive performance in forecasting the "HiNet Point Card" data. This analysis method can be used to determine feasible sales targets, or in the construction of a sales forecasting model.