自去年COVID-19 疫情實施三級警戒開始後，為了減少購買餐點時的人員接 觸， 符合食安規範的智能販賣機開始受到消費者的重視與青睞。智販機主要是 以「格子」的方式提供服務, 因此，格子的使用率(翻格率) 及對於銷售量的把控 是智販機營運成敗的關鍵。而機器的格子數量已確定，若不放滿，則可能損失 潛在利潤，但放太多可能賣不完，並基於食安考量，過一定時間即得下架。 因 此，提高使用率及精準把控上架商品的數量是智販機極為重要的營運目標。
想提升使用率，除了被動等待消費者到智販機前主動消費，若能提供線上銷 售通路，例如透過App 主動與消費者直接與智販機互動，這將有別於以往智販 機的銷售模式，並提升智販機的使用率。因此，本篇論文提出一個智慧販賣機 系統，具有線上銷售通路及後台維運的功能。不僅擴大銷售通路，也讓操作員 的工作流程簡化與程式化控制食品上架的問題，我們將其稱為智能販賣機管理 系統。此系統將符合現今軟體的設計架構，並提供Web、APP平台 實際導入商 業運用中，蒐集實際銷售資訊並予以優化銷售策略。
其中銷售策略最重要的部分即為上架商品的數量，精準的預測隔日銷售量 可以避免供貨少時而少賺的利潤，避免供貨太多未銷售完的浪費。藉由蒐集的 資料，透過機器學習的方法，如迴歸分析、時間序列的角度、與深度學習等模 型來建立模型，並使用平均絕對誤差、平均絕對百分比誤差、與均方根誤差來 衡量模型的好壞。我們與好食機合作，以在南港的智慧販賣機為實驗案例，蒐集從 2021/11 至 2022/04 的資料來建構模型，並用其模型來拿預測 2022/05 至 2022/06 的銷售量，並與實際銷量對比。實驗結果顯示，相較於使用 5 日移動 平均，每個月我們的系統可以幫好食機減少3.7 % 的損失

After the third level of COVID-19 alert, in order to reduce human contact when purchasing meals, vending machines that meet food safety specifications have been valued and favored by consumers. The vending machines mainly provide services in the form of ”grids”. The utilization rate of the grids (flipping rate) and the control of sales volume are the keys to the success or failure of the vending machine’s operation. The grid number of the machine has been determined already. Under the consideration of food safety, it’s a waste to put too much food. The food could not be sold after a certain period. Therefore, increasing the utilization rate and controlling the number of products on the shelves accurately are the critical goals of the vending machine.
Not only passively waiting for consumers to purchase at the vending machine, but we can also provide an online sales channel and allow consumers to directly interact with vending machines through the App to increase the utilization rate. It will change the sales model of vending machines which we are used to and increase the utilization rate of vending machines. This thesis proposes an in- formation system for the online sales channel and backend operations system of vending machines. It expands the sales channel, simplifies the operator’s workflow, and controls the issue of food restocking systematically. We call it an intelligent vending machine sales management system. This system will conform to the de- sign structure of software nowadays and will provide Web and App platforms that are imported into commercial applications to collect actual sales information so that we can optimize the sales strategies.
Among them, the most important part of the sales strategy is the volume of products on the shelves in the future days. Accurately predicting the sales volume of the next day can avoid losing profit due to the less supply, and also the loss due to excess supply. Our system is built with the collected data and machine learning methods such as regression analysis, time-series perspective, and other deep learning methods. The quality of the models is estimated by using the MSE (mean absolute error), MAPE (mean absolute percentage error), and RMSE (root mean square error). We cooperate with Howsfood.Inc to take sales status by the vending machines in Nangang as an experimental case and collect the data from 2021/11 to 2022/04 to construct a model and use it to predict the sales amount from 2022/05 to 2022/06. The result shows that we can help Howsfood.Inc reduces 3.7% loss if using our system against using the 5 MA method.

Abstract i
摘要 iii
致謝 v
Contents vi List of Figures ix List of Tables xi
1 Introduction 1
2 Related Work 5
2.1 VendingMachines............................ 5
2.2 Microservice............................... 7
2.3 ARIMA ................................. 9
2.4 LSTM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.5 LinearRegression............................ 12
3 System Design 14
3.1 VendingMachineManagementSystem ................ 14
3.1.1 BusinessScenario........................ 14
3.1.2 SystemOverview ........................ 16
3.1.3 ServiceWorker ......................... 17
3.1.4 DatabaseDesign ........................ 18
3.1.5 SystemImplementation .................... 19
3.2 SalesPredictionSystem ........................ 22
3.2.1 SystemOverview ........................ 23
3.2.2 SystemImplementation .................... 24
4 Model Design 25
4.1 MovingAverge ............................. 26
4.2 ARIMA ................................. 26
4.2.1 FindingParameters....................... 26
4.3 LSTM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3.1 ModelArchitecture....................... 28
4.3.2 ModelTraining ......................... 28
4.4 LinearRegression............................ 29
4.4.1 DecideIndependentVariables ................. 29
4.4.2 ConstructModel ........................ 32
4.4.3 RegressionModelTestandAnalysis. . . . . . . . . . . . . . 33
5 Evaluation Results 35
5.1 VendingMachineManagementSystem ................ 35
5.2 SalesPrediction............................. 39
5.2.1 MovingAverage......................... 39
5.2.2 ARIMA ............................. 41
5.2.3 LSTM.............................. 43
5.2.4 LinearRegression........................ 44
6 Conclusion and Future Work 46
Reference 48

[1] medRxiv. Covid-19 sars-cov-2 preprints from medrxiv and biorxiv. 2020.
[2] Canan Ag ̆alar and Derya O ̈ztu ̈rk Engin. Protective measures for covid-19 for healthcare providers and laboratory personnel. Turkish journal of medical sciences, 50(9):578–584, 2020.
[3] Department of statistics. https://dmz26.moea.gov.tw/GA/home/Home. aspx. Accessed: 2022-07-01.
[4] Colin C Venters, Rafael Capilla, Stefanie Betz, Birgit Penzenstadler, Tom Crick, Steve Crouch, Elisa Yumi Nakagawa, Christoph Becker, and Carlos Carrillo. Software sustainability: Research and practice from a software ar- chitecture viewpoint. Journal of Systems and Software, 138:174–188, 2018.
[5] Karandeep Singh, PM Booma, and Umapathy Eaganathan. E-commerce system for sale prediction using machine learning technique. In Journal of Physics: Conference Series, volume 1712, page 012042. IOP Publishing, 2020.
[6] N Cholianawati, WE Cahyono, A Indrawati, and A Indrajad. Linear regres- sion model for predicting daily pm2.5 using viirs-snpp and modis-aqua aot. In IOP Conference Series: Earth and Environmental Science, volume 303, page 012039. IOP Publishing, 2019.
[7] Ni Guo, Wei Chen, Manli Wang, Zijian Tian, and Haoyue Jin. Appling an improved method based on arima model to predict the short-term electricity consumption transmitted by the internet of things (iot). Wireless Communi- cations and Mobile Computing, 2021.
[8] Khaula Qadeer, Wajih Ur Rehman, Ahmad Muqeem Sheri, Inyoung Park, Hong Kook Kim, and Moongu Jeon. A long short-term memory (lstm) net- work for hourly estimation of pm2.5 concentration in two cities of south korea. Applied Sciences, 10(11):3984, 2020.
[9] Toshio Yokouchi. Today and tomorrow of vending machine and its services in japan. In 2010 7th International Conference on Service Systems and Service Management, pages 1–5. IEEE, 2010.
[10] Vending machine market in japan: Key research findings 2017. https:// www.yanoresearch.com/press/pdf/1795.pdf. Accessed: 2022-07-01.
[11] Ant’onio Raposo, Conrado Carrascosa, Esteban P’erez, Pedro Saavedra, Es- ther Sanju’an, and Rafael Mill’an. Vending machines: Food safety and quality assessment focused on food handlers and the variables involved in the indus- try. Food Control, 56:177–185, 2015.
[12] Ant’onio Raposo, Conrado Carrascosa, Esteban P’erez, Ana Tavares, Esther Sanju’an, Pedro Saavedra, and Rafael Mill’an. Vending machine foods: Eval- uation of nutritional composition. Italian journal of food science, 28(3), 2016.
[13] Feng-Cheng Lin, Hsin-Wen Yu, Chih-Hao Hsu, and Tzu-Chun Weng. Recom- mendation system for localized products in vending machines. Expert systems with applications, 38(8):9129–9138, 2011.
[14] Hideki Takei, Torey Hewitt, Michael Bantog, and Sana Becker. Evolutional dynamism and theoretical model of environmental and operational transfor- mation in vending machine retailing in usa and japan. Business Management and Strategy, 2(1):1, 2011.
[15] Hui Kang, Michael Le, and Shu Tao. Container and microservice driven design for cloud infrastructure devops. In 2016 IEEE International Conference on Cloud Engineering (IC2E), pages 202–211. IEEE, 2016.
[16] Martin Fowler microservices. https://martinfowler.com/articles/ microservices.html. Accessed: 2022-07-01.
[17] Deval Bhamare, Mohammed Samaka, Aiman Erbad, Raj Jain, and Lav Gupta. Exploring microservices for enhancing internet qos. Transactions on Emerging Telecommunications Technologies, 29(11):e3445, 2018.
[18] Kapil Bakshi. Microservices-based software architecture and approaches. In 2017 IEEE aerospace conference, pages 1–8. IEEE, 2017.
[19] Vinod Kumar Vavilapalli, Arun C Murthy, Chris Douglas, Sharad Agarwal, Mahadev Konar, Robert Evans, Thomas Graves, Jason Lowe, Hitesh Shah, Siddharth Seth, et al. Apache hadoop yarn: Yet another resource negotiator. In Proceedings of the 4th annual Symposium on Cloud Computing, pages 1–16, 2013.
[20] Benjamin Hindman, Andy Konwinski, Matei Zaharia, Ali Ghodsi, Anthony D Joseph, Randy Katz, Scott Shenker, and Ion Stoica. Mesos: A platform for {Fine-Grained} resource sharing in the data center. In 8th USENIX Sympo- sium on Networked Systems Design and Implementation (NSDI 11), 2011.
[21] Abhishek Verma, Luis Pedrosa, Madhukar Korupolu, David Oppenheimer, Eric Tune, and John Wilkes. Large-scale cluster management at google with borg. In Proceedings of the Tenth European Conference on Computer Systems, pages 1–17, 2015.
[22] Kubernetes. https://kubernetes.io. Accessed: 2022-07-01.
[23] Maobin Li, Shouwen Ji, and Gang Liu. Forecasting of chinese e-commerce sales: an empirical comparison of arima, nonlinear autoregressive neural net- work, and a combined arima-narnn model. Mathematical Problems in Engi- neering, 2018.
[24] Adebiyi A Ariyo, Adewumi O Adewumi, and Charles K Ayo. Stock price prediction using the arima model. In 2014 UKSim-AMSS 16th international conference on computer modelling and simulation, pages 106–112. IEEE, 2014.
[25] Klaus Greff, Rupesh K Srivastava, Jan Koutn ́ık, Bas R Steunebrink, and Ju ̈rgen Schmidhuber. Lstm: A search space odyssey. IEEE transactions on neural networks and learning systems, 28(10):2222–2232, 2016.
[26] Paul J Werbos. Backpropagation through time: what it does and how to do it. Proceedings of the IEEE, 78(10):1550–1560, 1990.
[27] Jian Qi Wang, Yu Du, and Jing Wang. Lstm based long-term energy con- sumption prediction with periodicity. Energy, 197:117197, 2020.
[28] Zunya Shi and Abdallah Chehade. A dual-lstm framework combining change point detection and remaining useful life prediction. Reliability Engineering & System Safety, 205:107257, 2021.
[29] H Roopa and T Asha. A linear model based on principal component analysis for disease prediction. IEEE Access, 7:105314–105318, 2019.
[30] T Gopalakrishnan, Ritesh Choudhary, and Sarada Prasad. Prediction of sales value in online shopping using linear regression. In 2018 4th International Conference on Computing Communication and Automation (ICCCA), pages 1–6. IEEE, 2018.
[31] Harlili Luminto. Weather analysis to predict rice cultivation time using multi- ple linear regression to escalate farmer’s exchange rate. In 2017 International Conference on Advanced Informatics, Concepts, Theory, and Applications (ICAICTA), Denpasar, Indonesia, pages 16–18, 2017.
[32] Yanming Yang. Prediction and analysis of aero-material consumption based on multivariate linear regression model. In 2018 IEEE 3rd international con- ference on cloud computing and big data analysis (icccbda), pages 628–632. IEEE, 2018.
[33] Google pub/sub. https://cloud.google.com/pubsub/docs/overview. Ac- cessed: 2022-07-01.
[34] Cronjob. https://kubernetes.io/docs/concepts/workloads/ controllers/cron-jobs/. Accessed: 2022-07-01.
[35] Ecpay. https://www.ecpay.com.tw, . Accessed: 2022-07-01.
[36] Ecpay sdk. https://github.com/ECPay/ECPayAIO_Node.js, . Accessed: 2022-07-01.
[37] Nishtha Jatana, Sahil Puri, Mehak Ahuja, Ishita Kathuria, and Dishant Go- sain. A survey and comparison of relational and non-relational database. International Journal of Engineering Research & Technology, 1(6):1–5, 2012.
[38] Google firebase. https://firebase.google.com. Accessed: 2022-07-01.
[39] Line login. https://developers.line.biz/zh-hant/services/
line-login. Accessed: 2022-07-01.
[40] Appcenter. https://appcenter.ms. Accessed: 2022-07-01.
[41] Adam Boduch and Roy Derks. React and React Native: A complete hands-on guide to modern web and mobile development with React. js. Packt Publishing Ltd, 2020.
[42] Node.js Express. https://expressjs.com/zh-tw/. Accessed: 2022-07-01.
[43] Container Registry — Google Cloud. https://cloud.google.com/
container-registry. Accessed: 2022-07-01.
[44] Tiobe programming community. https://www.tiobe.com/tiobe-index/.
Accessed: 2022-07-01.
[45] Lee Sechrest and Souraya Sidani. Quantitative and qualitative methods:: Is
there an alternative? Evaluation and program planning, 18(1):77–87, 1995.
[46] Jinah Kim and Nammee Moon. A deep bidirectional similarity learning model using dimensional reduction for multivariate time series clustering. Multime- dia Tools and Applications, 80(26):34269–34281, 2021.
[47] Autoregressive integrated moving average (arima). https://www. investopedia.com/terms/a/autoregressive-integrated-moving-average-arima. asp. Accessed: 2022-07-01.
[48] Eric-Jan Wagenmakers and Simon Farrell. Aic model selection using akaike weights. Psychonomic bulletin & review, 11(1):192–196, 2004.
[49] Thabani Nyoni. Box-jenkins arima approach to predicting net fdi inflows in zimbabwe, munich university library – munich personal repec archive (mpra), paper no. 87737. 2018.
[50] The unreasonable effectiveness of recurrent neural networks. http:// karpathy.github.io/2015/05/21/rnn-effectiveness/. Accessed: 2022- 07-01.
[51] Dokkyun Yi, Jaehyun Ahn, and Sangmin Ji. An effective optimization method for machine learning based on adam. Applied Sciences, 10(3):1073, 2020.
[52] Andrius Vabalas, Emma Gowen, Ellen Poliakoff, and Alexander J Casson. Machine learning algorithm validation with a limited sample size. PloS one, 14(11):e0224365, 2019.
[53] Central Weather Bureau. https://www.cwb.gov.tw/V8/C/. Accessed: 2022- 07-01.
[54] Iryna A Sievidova et al. Factors affecting the economic management effi- ciency of agricultural enterprises in ukraine. Problems and Perspectives in Management, 15(2 (c. 1)):204–211, 2017.