由於無線網路是透過廣播的方式傳遞資料，人們很容易的透露出自己的個人資料，包含關於隱私的資料。然而, 我們也可以透過這些無線網路的資料，來讓我們更加了解每個人的生活風格與個性. 我們利用時間與空間性的分析來分析顧客，而這些資訊當中並不會包含各人隱私有關的資料。

隨著物聯網的發展，生活中隨處可見各式各樣的感應裝置在追蹤人們的生活狀況，而無線網路的蓬勃發展加速了此進程，如何觀察顧客與店家的互動方法更加多元。由於智慧型裝置的普遍，顧客的行為更多的透過手上的裝置來完成，不單純的只是打電話，更可透過無線網路來進行社交、看影片之行為。

此篇論文以分析顧客無線網路資訊，我們在咖啡廳實地收集店所發出的無線網路訊號裡。在咖啡廳當中架設一台AP，並利用AP裝置監聽廣播在空間中的訊號，收集連線要求與資料要求之訊號，並將之即時回傳至儲存裝置當中。而這些訊號並不包含任何個人隱私有關的資訊，我們透過對這些資料的時間與空間的來分析，利用分析結果來做分群，以便於適應不同顧客群體的習慣。利用建置多個微型模型來達到的預測同時得到與顧客個人有關之資訊。我們利用支援向量回歸透過對於過去資料的訓練，來預測未來顧客來店人數。同時我們也使用支援向量分類器來預測顧客來店與否，顧客來店之後其內用外帶之判斷。

Becasue of the transmitting of wireless network using broadcasting, people easily leak their personal information, including the privacy information. However, due to the information from public wireless, it's easy for us to understand people's life style and personality. We use the spatial and temporal information to analysis the customers without fetching any personal privacy.

As the development of IoT, lots of kind and number of sensing deveces monitoring human life, wireless network increase the speed of the procedure. people rely on their smart devices to communicate, watch video, play games, and so on. That's why we can provide shop owners different kind of information by observing and analysis different kind of customers. With these information, the shop can give customized service.

This thesis analysis the Wi-Fi information of customers, we collected Wi-Fi signal in the real world environment coffee shop. We seted a AP in the coffee shop, use it to listen the signal that devices broadcasted in the environment, that included connection request and data request signals, then we save those data into the storge in real time. Those information doesn't include personal privacy related information.

We use this data to do spatial and temporal analysis, and use the results clustering the custoemrs to suit different kind of habbit group of customers. We build the models for each cluster called micro models to predict the information related to the customers. We use SVR trained by historical data to predict the number of coming customers in the future. And we also use SVM to predict the customers will come to the coffee shop or not. If the customer comes in, we predict he/she will be stay-in or take-out.

[1] Z. Fu, M. Almgren, O. Landsiedel, and M. Papatriantafilou, “Online temporal-spatial analysis for detection
of critical events in cyber-physical systems,” in Big Data (Big Data), 2014 IEEE International
Conference on, pp. 129–134, IEEE, 2014.
[2] F. Chen, J. Dai, B. Wang, S. Sahu, M. Naphade, and C.-T. Lu, “Activity analysis based on low sample
rate smart meters,” in Proceedings of the 17th ACM SIGKDD international conference on Knowledge
discovery and data mining, pp. 240–248, ACM, 2011.
[3] S. Emtiyaz and M. Keyvanpour, “Customers behavior modeling by semi-supervised learning in customer
relationship management,” arXiv preprint arXiv:1201.1670, 2012.
[4] S. Goel, J. M. Hofman, S. Lahaie, D. M. Pennock, and D. J. Watts, “Predicting consumer behavior with
web search,” Proceedings of the National academy of sciences, vol. 107, no. 41, pp. 17486–17490,
2010.
[5] S. Jiang, J. Ferreira Jr, and M. C. Gonzalez, “Discovering urban spatial-temporal structure from human
activity patterns,” in Proceedings of the ACM SIGKDD international workshop on urban computing,
pp. 95–102, ACM, 2012.
[6] Z. Suhong, Y. Lijun, and D. Lifang, “The spatial-temporal pattern of people’s daily activities and transportation
demand analysis-a case study of guangzhou, china,” in Management and Service Science
(MASS), 2010 International Conference on, pp. 1–4, IEEE, 2010.
[7] C.-C. Chang and C.-J. Lin, “LIBSVM: A library for support vector machines,” ACM Transactions
on Intelligent Systems and Technology, vol. 2, pp. 27:1–27:27, 2011. Software available at http:
//www.csie.ntu.edu.tw/~cjlin/libsvm.