《东海大学统计学系硕士班》由会员分享，可在线阅读，更多相关《东海大学统计学系硕士班（69页珍藏版）》请在金锄头文库上搜索。
1、東海大學統計學系碩士班東海大學統計學系碩士班 碩士論文碩士論文 指導教授：林雅俐指導教授：林雅俐 博士博士 以以B為依據為依據之之旅館旅館住宿滿意度住宿滿意度資料資料採礦採礦 Satisfaction of Hotel Accommodation using Statistical Data Mining on B 研究生：研究生：葉樺蓁葉樺蓁 中華民國中華民國 一百零一百零四四年年 六六 月月 中文摘要中文摘要 為了方便旅客藉由訂房網站快速決定住宿地點，訂房網站業者邀請過去在網站訂房且有入住者就整潔度、舒適程度、飯店地點、設施、員工素質、以及性價比等指標提供住房評價，將整體評分呈現於網頁，做
2、為潛在旅客住宿之選擇參考。本研究使用B網站所提供之689家臺灣地區旅館，超過七萬名旅客之評價分數，探討影響旅客住宿選擇的關鍵因素，並利用多元迴歸模型、決策樹模型、和類神經網路模型等資料採礦方法，建立整體評分之最佳預測模型。研究結果發現，線性迴歸模型、決策樹模型、類神經網路模型皆以整潔度和飯店地點為預測整體評分之關鍵因素，迴歸模型之解釋能力可高達93.6%，決策樹模型之平均平方誤差為0.38，類神經網路模型之平均平方誤差為0.1。進一步將整體評分分為7分以上和未滿7分兩類，決策樹模型和類神經網路模型皆顯示整潔度與設施分別對提升整體評分上有顯著的影響力，在決策樹模型方面，整潔度的預測錯誤分類比率為
3、0.094，設施的預測錯誤分類比率為0.08，在類神經網路模型方面，整潔度和設施的預測錯誤分類比率皆為0.094。由以上的結果發現，在預測整體評分模型中，整潔度和飯店地點是旅客所在乎的因素，而在分類模型中，則是整潔度和設施，但考量成本，改善整潔度所帶來的效益會高於設施，並可以在短時間內提高整體評分。因此整潔度和飯店地點是影響旅客住宿選擇的關鍵因素。 關鍵字：住宿選擇、資關鍵字：住宿選擇、資料探勘、決策樹模型、類神經網路模型料探勘、決策樹模型、類神經網路模型 Abstract In order to let tourists decide the accommodation location q
4、uickly with the booking websites, website owners invite those who had booked hotel via internet , lived there and evaluated with benchmarks such as cleanness, comfort, locations, facilities, staff and value for money. The review score presented on the website are provided as the recommendation of ch
5、oosing hotels for potential travelers. Using the data provided by B website which has evaluation scores given by over seventy thousand visitors in 689 Taiwan hotels, this research will explore the key factors of affecting the tourists accommodation options. By using the data mining methods like mult
6、iple line regression, decision tree models and neural network models, we could establish the prediction model which best fit to the review score. The results indicate both cleanness and locations are the key factors for the review score. The explain review score of the regression model can be up to
7、93.6%, the average square error of the decision tree model was 0.38, and the average square error of the neural network model is 0.1. Furthermore, we divide review score into two types as above 7 and below 7. Besides, according to decision tree model and neural network model, we discover that cleann
8、ess and facility have significant influence in elevating review score. In the aspect of decision tree, misclassification of cleanness and facilities are 0.094 and 0.08. But, in neural network, the misclassification of both two neural network models are 0.094. By examining the earlier result, in a pr
9、ediction model of review score, we could conclude that cleanness and locations are the factors that tourists care about, and in a model of classification, it was the cleanness and facilities. However, from the cost, improving the cleanness will give more benefits than upgrading the facilities, and i
10、t could increase the review score in short amount of time. Key-words: Booking Select, Data Mining, Decision Tree Model, Neural Network Model 目錄目錄 第一章 緒論 . 1 1.1 研究動機 . 1 1.2 研究目的 . 2 1.3 研究架構 . 3 第二章 文獻探討 . 4 2.1 住宿選擇 . 4 2.2 住宿滿意度 . 5 第三章、研究方法 . 7 3.1 資料來源 . 7 3.2 變數說明 . 8 3.3 決策樹模型 . 10 3.4 類神經網路模型 . 12 第四章 研究結果 . 14 4.1 信度與效度分析 . 14 4.2 敘述性統計 . 15 4.3 迴歸模型建構 . 22 4.4 決策樹模型 .
