醫藥大數據－應用多元時間序列分析探討網路輿論聲量及乾眼症用藥量與隱形眼鏡銷售量之關聯研究

Abstract

近十年來，我國乾眼症患者有低齡化的趨勢，盛行率已從10-15%攀升至25%， 而造成乾眼症的症狀為數種原因共同作用下的結果，然而主要還是來自於現代人生活型態的改變(馬搖駸，2017)，如屏幕顯示終端的頻繁使用、隱形眼鏡的配戴，又諸如外部環境因素，如較常使用私人運具騎乘致使眼部接觸空氣髒汙等，若能及早透過生活習慣的調整及選用合適的眼部醫療器材，將延緩乾眼症的發生。 本研究欲透過高雄市之乾眼症與隱形眼鏡配戴造成角膜水腫之用藥量，及網路社群中有關隱形眼鏡商品及配戴之主題式討論聲量，探討兩者與隱形眼鏡銷售量之關聯及影響程度。本研究之研究目的在於從用藥量之資料觀察用藥行為模式，從網路討論聲量歸納出對於隱形眼鏡挑選及衍生病症的關注度，進一步與現代人用眼、配戴隱形眼鏡的生活習慣作連結，了解不同情境下，消費者適合之隱形眼鏡種類，提供隱形眼鏡業者制定銷售策略的憑據，使銷售策略效益最大化。 本研究採用多元時間序列分析法，對標的藥品用藥量及網路社群主題式討論聲量與隱形眼鏡銷售量之關聯建立多元時間序列模型，於建立模型過程中，首先對三者時間序列資料進行平穩性檢驗，藉由觀察序列資料之自相關及偏相關函數圖，來決定ARIMAX模型之自迴歸及移動平均參數；模型擬合過後，對其估計之參數係數進行顯著與否的檢定，對自變數和依變數之間的關聯使用格蘭傑因果關係檢驗，並檢驗模型之殘差序列是否平穩且不具高度自相關，使得模型的精確度不受殘差影響。最後，以依據信息準則挑選之優良模型，探討標的藥品用藥量及網路社群主題式討論聲量之變化對隱形眼鏡銷售量之影響程度，並作為未來掌握銷售量趨勢之基礎，建立帶來營收效益的銷售策略。

In recent years, the young population with dry eye syndrome increases in Taiwan, and the prevalence rate has increased from 10-15% to 25%. The symptoms of dry eye is caused by several factors, but mainly because of the modern lifestyle changes, such as frequently using of visual display terminal, contact lens wearing, but also external environmental factors. If people can adjust living habits and use contact lens appropriately, the dry eyes syndrome will not occur early. This study investigates the relationship among medicine dosage of dry eye syndrome and corneal bedewing caused by contact lens wearing in Kaohsiung City, the topical discussion of contact lens products in online community and the contact lens’ sales volume and the degree of influence, and to further analyzing their link with people’s habits of wearing contact lenses. This study provides contact lens manufacturers with evidence to maximize the effectiveness of sales strategies. In this study, a multivariate time series analysis method is used to establish a multivariate time series model of the target medicine dosage and the online community discussion’s influence on the volume of contact lens sales. In the process of establishing the model, the stationarity test is performed. The autoregressive and partial correlation function plots are observed to determine the autoregressive and moving average parameters of the ARIMAX model. The excellent model is selected according to the information criteria, to discuss the change of target medicine dosage and topical discussion of online community's impact on sales of contact lenses, and to plan efficient sales strategy based on the model.