此篇研究我們主要參考UTAUT理論，探討如何預測一個使用者是否會在SaaS上付費。在過去的研究當中，大多數的學者專注於研究什麼樣的因素能夠解釋使用者的購買意圖或是行為，因此傳統的IS研究領域被數據分析的學者們認為是“因果的解釋”，而不是“進一步預測”。然而，致力於數據分析的學者們，通常會收集大量的資料去預測使用者是否會付費，但是這樣的作法缺乏了理論上的架構，而理論的架構正是IS領域的人們相當在乎的。因為上述所提及的矛盾，這篇研究提出一個以理論為基礎的預測方法，此篇研究的資料取自一篇典型、因果解釋的IS研究：“從免費到付費:免費增值模式下SaaS用戶的使用者行為研究。”（蔡偉志, 2015），來執行以理論為基礎的預測。另外，我們也額外加入了問卷網站的系統後台資料，做了一個純粹數據分析的預測。最後我們比較這兩個分析：（1）以理論為基礎的預測（2）純粹數據分析的預測。我們會從目標定義、資料準備一直到最後產出最終模型，逐步地比較兩者，探討兩者不同之處與各自做了哪些妥協。
總體來說，此篇研究的結果顯示在以理論為基礎的預測當中，最好的一個預測模型擁有77.99%的整體預測正確率；而在純粹數據分析中的逐步回歸有77.61%的正確率，這說明以理論為基礎的預測不亞於純粹數據分析的預測。另外，若是只針對付費的預測準確率，以理論為基礎的預測是52.86%，純粹數據分析的預測是43.66%，這說明以理論為基礎的預測更勝於純粹數據分析的預測，勝於近10%。因此有理論的幫助下，仍然有預測的能力。此篇研究提供給那些想要做預測同時又想保留理論架構的人們，一個明確的指引。


In this study, we investigate how to predict whether a user will pay for SaaS under UTAUT theory. In previous studies, scholars in IS field usually focus on what kind of factors explain users’ buying intension or behavior. Hence, it is said that traditional IS studies are mainly “causal explanation” rather than “further prediction”. On the other hand, scholars doing pure prediction usually collect tons of data to predict if users will pay or not, but this approach lack of a theoretical framework which people in IS field care much about. Because of these contradictions, this study proposes a theory-based prediction. We take the data from a thesis “From Free Using to Paying: Differences Between Users’ Attitude and Behavior in SaaS Freemium Model ” (Tsai, 2015), a typical causal explanation IS research using UTAUT theory, to do a theory-based prediction. Furthermore, we add additional meta-data to do a pure data-mining prediction. In the end, we compare two analyses: (1) a theory-based prediction (2) a pure data mining prediction. The comparison goes step by step starting from goal definition, data preparation, all the way to model report. In each step, we discuss the differences and how the compromise heed to be made while doing each analysis.
Overall, the results of this study shows that one of the best model from theory-based prediction got 77.99% of overall accuracy rate and the one of stepwise from pure data-mining prediction is 77.61%. It shows that theory-based prediction is not worse than pure data-mining prediction. As for the paying accuracy rate is 52.86% and 43.66%, separately. It shows that theory-based prediction is even better than pure data mining prediction and is almost 10% higher than pure data mining prediction. Therefore, with theory, there is still predictive power. The study provides a practical guidance for people who want to do a prediction and retain the theoretical framework at the same time.

Chapter 1 - Introduction P.1
Chapter 2 - Literature review P.4
2.1. Explanatory modeling and predictive modeling P.4
2.1.1. Explanatory modeling P.4
2.1.2. Predictive modeling P.6
2.2. Research gap P.7
2.3. Theoretical foundation P.7
2.3.1. UTAUT P.8
2.3.2. Research model P.9
Chapter 3 - Methodology P.12
3.1. Goal definition P.12
3.1.1. Business goal P.12
3.1.2. Analytics goal P.12
3.2. Study design and data collection P.12
3.3. Data preparation P.15
3.3.1. Missing value handling P.15
3.3.2. Data partition P.15
3.4. Exploratory data analysis(EDA) P.16
3.5. Choice of variables P.17
3.6. Choice of methods P.18
3.7. Model evaluation, validation and selection P.19
3.8. Model use and reporting P.19
Chapter 4 - Results of theory-based prediction P.20
4.1. Empirical result of of path analysis P.20
4.2. Logistic regression model P.21
4.3. Empirical results of theory-based prediction P.23
4.3.1. Over-fitting P.23
4.3.2. Accuracy rate P.23
4.3.3. Decile lift chart P.25
Chapter 5 - Results of pure data mining prediction P.28
5.1. Pure data mining prediction P.28
5.1.1. Propose of comparison P.28
5.1.2. Detailed of analysis P.28
5.1.3. Statistical summaries and comparison P.31
5.2. A different random seed P.35
5.2.1. Propose of comparison P.35
5.2.2. Detailed of analysis P.35
5.2.3. Statistical summaries and comparison P.35
Chapter 6 - Conclusion P.39
6.1. Contribution P.39
6.2. Limitation and recommendation for further research P.40
Reference P.41
Appendix A P.43
Appendix B P.47

Chu, C. W., & Lu, H. P. (2007). Factors influencing online music purchase intention in Taiwan: An empirical study based on the value-intention framework. Internet Research, 17(2), 139-155.
Cortez, P., & Morais, A. D. J. R. (2007). A data mining approach to predict forest fires using meteorological data.
Doty, D. H., & Glick, W. H. (1994). Typologies as a unique form of theory building: Toward improved understanding and modeling. Academy of management review, 19(2), 230-251.
Fetscherin, M., & Lattemann, C. (2008). User acceptance of virtual worlds. Journal of electronic commerce research, 9(3), 231.
Gefen, D., Karahanna, E., & Straub, D. W. (2003). Trust and TAM in online shopping: an integrated model. MIS quarterly, 27(1), 51-90.
Gregor, S. (2006). The nature of theory in information systems. MIS quarterly, 611-642.
Ho, C. F., & Wu, W. H. (1999, January). Antecedents of customer satisfaction on the Internet: an empirical study of online shopping. In Systems Sciences, 1999. HICSS-32. Proceedings of the 32nd Annual Hawaii International Conference on (pp. 9-pp). IEEE.
López-Nicolás, C., Molina-Castillo, F. J., & Bouwman, H. (2008). An assessment of advanced mobile services acceptance: Contributions from TAM and diffusion theory models. Information & Management, 45(6), 359-364.
Lu, H. P., & Hsiao, K. L. (2010). The influence of extro/introversion on the intention to pay for social networking sites. Information & Management, 47(3), 150-157.
Putka, D. J., Beatty, A. S., & Reeder, M. C. (2016). Modern prediction methods: New perspectives on a common problem. Organizational Research Methods, 1094428117697041.
Porter, M. E. (1979). How competitive forces shape strategy.
Rhee, H. S., Kim, C., & Ryu, Y. U. (2009). Self-efficacy in information security: Its influence on end users' information security practice behavior. Computers & Security, 28(8), 816-826.
Rosendaal, F. R. & Reitsma, P. H. (2015). Prediction. Journal of Thrombosis and Haemostasis, Vol.13(10), pp.1741-1741
Semenzin, D., Meulendijks, E., Seele, W., Wagner, C., & Brinkkemper, S. (2012, June). Differentiation in Freemium: Where Does the Line Lie?. In International Conference of Software Business (pp. 291-296). Springer Berlin Heidelberg.
Shmueli, G. (2010). To explain or to predict?. Statistical science, 25(3), 289-310.
Shmueli, G., & Koppius, O. R. (2011). Predictive analytics in information systems research. Mis Quarterly, 553-572.
Shmueli, G., Patel, N. R., & Bruce, P. C. (2016). Data Mining for Business Analytics: Concepts, Techniques, and Applications with XLMiner. John Wiley & Sons.
Shmueli, G., Ray, S., Estrada, J. M. V., & Chatla, S. B. (2016). The elephant in the room: Predictive performance of PLS models. Journal of Business Research, 69(10), 4552-4564.
Wang, Y. S., Yeh, C. H., & Liao, Y. W. (2013). What drives purchase intention in the context of online content services? The moderating role of ethical self-efficacy for online piracy. International Journal of Information Management, 33(1), 199-208.
Wagner, T. M., Benlian, A., & Hess, T. (2014). Converting freemium customers from free to premium—the role of the perceived premium fit in the case of music as a service. Electronic Markets, 24(4), 259-268.
Westland, J. C., & Clark, T. H. (2001). Global electronic commerce theory and case studies. Universities Press.