Forecasting Product Styles with Time Series Models

• Main Authors: Zhang Zhen-Zhen, 張珍珍
• Other Authors: Chang Chien-Cheng
• Format: Others
• Language: en_US
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/02500218980317724413

碩士 === 華梵大學 === 工業設計學系碩士班 === 96 === Competition of the product market is becoming more intense. There are successful market opportunities when the products satisfy the customers’ requirements. Therefore, it is essential to sustain the performance of a product. In this study, consumer’s performances were toward specific types of mobile phones at different periods of time. Twenty-two black and white mobile phones samples and nine image pairs were used in the product image evaluation experiment. A seven-pointed semantic scale was used for the measurement of user’s perceptions toward mobile phone sample images. The effect of different time intervals on user’s preference and image perceptions to product images was explored. In the study, the author used cluster analysis to specify groups of samples by the Ward’s method from which three or four groups were specified for the samples evaluated in the experiment. Then, nonparametric test (Wald-Wolfowitz test – Runs test), under different time intervals was conducted to determine whether the representative samples and pairs came from the same type of product styles. Finally, time series analysis regarding each group of pairs, war performed to determine the most suitable ARIMA model for the prediction of preference and product style perceptions. We hope to provide concrete references for mobile phones design in the future. The result showed no significant differences in the perceptions of handy, novel, compact, modern, delicate, feminine, familiarity, preference and purchase intention at different time intervals. From cluster analysis, twenty-two samples were divided into 3 ~ 4 groups in nine image words. From the result of time series analyses, twenty-eight best ARIMA models and twenty-eight prediction equation were identified. Among them Handy MA(1), MA(1), MA(2); Novel MA(1), AR(2), MA(1); Compact AR(3), AR(1), ARIMA(3,0,1); Modern MA(1), AR(3), MA(1); Delicate ARIMA(3,0,2), AR(1), ARIMA(2,0,1); Feminine MA(2), ARIMA(3,0,1), MA(1), MA(1); Familiarity MA(1), AR(1), MA(1), MA(1); Preference MA(2), MA(1), MA(1); Purchase Intention AR(1), MA(1), MA(1) could be selected to determine the best fitting forecasting model for the image words of representative sample. From Box-Ljung statistics, the model residuals presented the white noise, meaning that these ARIMA model made most reliable forecasts.