天氣因素於網路團購消費者行為之影響研究

The Relationship of Online Group Buying Behavior Incorporating the Effect of Weather

本研究目的在驗證天氣因素為網路團購消費者之感知有用性之一，並補充網路團購消費者感知的相關研究。本研究利用台灣網路團購的真實資料及氣象觀測資料，運用機器學習中四種分類模型進行比較，首先實證預測模型，再進行天氣預測感知分析。研究結果顯示，在十大產品類別中，隨機森林(Random Forest)為實證中較優模型；其次，在六種天氣因素中，最重要之天氣變數是體感溫度和日照時數。研究結果印證天氣變化影響消費者之評價，與既有研究(Chu et al., 2013；Zwebner et al., 2014；Watanabe et al., 2016；Tian et al., 2018；Schlager et al., 2020)呼應，體感溫度為本研究之創見。再者，體感溫度不僅為重要之天氣變數，體感溫度之高低也影響不同地區、品類之銷售，再次說明團購消費者行為與天氣具有關聯。

The purpose of this research is to verify that the weather factor is one of the perceived usefulness of online group buying consumers, and to supplement related research on the perception of online group buying consumers. This study uses one of Taiwan's online group real data and uses the meteorological observation data to supplement the related research in Taiwan. We test four models of machine learning to find out the best predictive model for online marketing and then conduct weather forecast analysis. These results show that among the ten product collections, random forest method is the empirical best practice, and among the six weather factors, the most important weather variables for the online group buying platform as a whole are apparent temperature and sunshine duration. Apparent temperature is the original idea of the research. These findings consent that online consumers behavior are effected by weather changes, and echo with existing researches (Chu et al., 2013；Zwebner et al., 2014；Watanabe et al., 2016；Tian et al., 2018；Schlager et al., 2020). The results also exhibit that the apparent temperature is not only an important weather variable, but also affects the sales of different locations and collections. This research contributes to the analysis of the diversity of group buying behavior and weather marketing.