臺灣博碩士論文加值系統

公司物料採購與庫存水準管理通常根據銷售數量進行安排，本研究探討扣件外銷公司面臨訂單與原物料特性變動下，以往的經驗法則無法有效平衡需求，導致供應鏈出現待料停機和延遲交貨等問題。研究提出一個結合量化與質性分析的銷售預測模型，採用混合型方式，將量化的公司歷史資料，進行差分整合移動平均自我迴歸模型（ARIMA）資料分析，並使用長短期記憶模型（LSTM）進行銷售預測，質性的專家報告利用自然語言處理（NLP）的預訓練模型微調分析，彙整多篇專家文章，以提供更準確的銷售預測和趨勢分析。結果顯示，該模型在疫情期間排除後，預測精準度以判定係數（R²）為0.837，顯示具有高度的實務可行性，可有效掌握未來銷售變動，使企業提前準備原物料，提高供應鏈韌性。本研究混合型方法克服了機器學習僅依賴歷史資料的限制，通過整合專家報告，能提前看見市場變動，提高銷售預測的準確性和能力，給企業充足的時間進行庫存整合和管理，保持市場韌性和競爭力。總結來說，研究結合量化預測模式與質性文章趨勢分析判斷，提供更全面性的未來銷售變化趨勢參考資訊。

Material procurement and inventory level management in companies are often arranged based on sales volume. This study explores the challenges faced by fastener export companies in balancing demand under order and raw material property changes. The traditional rule of thumb cannot effectively balance demand, leading to supply chain problems such as waiting for materials and delayed delivery. This study proposes a sales forecasting model that combines quantitative and qualitative analysis. It adopts a hybrid approach to analyze quantitative corporate historical data using the difference integration moving average self-regressive model (ARIMA) and long short-term memory (LSTM) for sales forecasting. Qualitative expert reports are analyzed using the fine-tuning analysis of the natural language processing (NLP) pre-training model, and multiple expert articles are compiled to provide more accurate sales forecasting and trend analysis. The results show that the prediction accuracy of the model is 0.837 in terms of the coefficient of determination (R²) after excluding the epidemic period, which shows that it has high practical feasibility and can effectively grasp future sales changes, allowing enterprises to prepare raw materials in advance and improve supply chain resilience. The hybrid method of this study overcomes the limitation that machine learning only relies on historical data. By integrating expert reports, it can anticipate market changes, improve the accuracy and ability of sales forecasting, and provide enterprises with sufficient time for inventory integration and management to maintain market resilience and competitiveness. In summary, the study combines quantitative forecasting mode with qualitative article trend analysis judgment to provide more comprehensive reference information on future sales change trends.

摘 要 i
ABSTRACT ii
誌 謝 iii
目 次 iv
表目錄 vi
圖目錄 vii
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 3
1.3 論文架構 4
第二章 文獻探討 6
2.1 台灣扣件產業的發展 6
2.2 銷售預測模式 8
2.2.1統計類型方法 8
2.2.2 人工智慧類型方法 10
2.3 自然語言處理與預訓練模型微調 14
2.4 混合式方法 18
第三章 研究方法 21
3.1 研究架構與流程 21
3.2 研究資料來源與前處理 23
3.2.1資料來源 23
3.2.2資料前處理 24
3.3 量化的銷售預測模型建構 26
3.3.1 ARIMA模型分析資料集特性 26
3.3.2 LSTM預測模型建構 30
3.3.3 預測結果評估方式 32
3.4 文章的BRET微調分類模型建構 33
3.4.1 設定標籤 33
3.4.1 BRET微調分類模型建構 34
3.4.2 多篇文章的分類彙整 36
3.5 混合型銷售預測模型 38
第四章 研究結果與討論 39
4.1 量化的銷售預測分析 39
4.1.1 ARIMA分析資料集時間序列特性 39
4.1.2 LSTM模型預測結果分析 42
4.1.3 LSTM預測結果評估指標 54
4.2 蒐集文章的分類預測分析 55
4.2.1 BERT模型微調的分類預測結果與分析 55
4.2.2 多篇文章的分類結果整合 61
4.3 混合型銷售預測結果分析 65
4.3.1 單月銷售預測應用說明 65
4.3.2 歷年資料驗證分析 69
4.3.3混合型銷售預測分析結果彙總 72
第五章 結論與未來展望 75
5.1 結論 75
5.2 未來研究方向 76
參考文獻 77

1.方柏超(2021年5月14日)。台灣產業發展與政府補助的相關性。myMKC管理知識中心。https://mymkc.com/article/content/24457
2.王瑜甄、吳軒萍、吳威廷、黃美文、詹浩玥、郭文嘉、鍾榮哲（2021）。後疫情時代製造業數位轉型。逢甲大學國際經營與貿易學系。
3.王禎祥（2021）。應用機器學習回歸模型於成衣製造業銷售預測-以T公司為例〔未出版之碩士論文〕。國立中央大學資訊管理學系。
4.呂駿彬（2020）。台灣扣件產業發展策略之探究〔未出版之博士論文〕。國立成功大學政治學系研究所。
5.宋文龍（2023）。風雨70年台灣扣件產業的軟實力。產業動態惠達雜誌，201，94-96。
6.林顯明（2019）。臺灣螺絲帽（扣件）產業發展與轉型之政治經濟分析。國會季刊，47(2)，49-82。
7.武小平、張強、趙芳、焦琳（2021）。基於 BERT 的心血管醫療指南實體關係抽取方法。電腦應用，41(1)，145-149。
8.況國安（2015）。台灣扣件產業轉型高值化之探討〔未出版之碩士論文〕國立中正大學高階主管管理。
9.紀翔瀛（2021年9月29日）。2020年我國扣件產業回顧與展望。工商時報。https://readers.ctee.com.tw/cm/20210929/a77asa7/1147478/share
10.紀翔瀛（2022）。2022上半年台灣扣件業的回顧與展望。惠達雜誌產業報導，196，62-65。https://fastener-world.com.tw/data/pdf_download/FW_196_C_62.pdf
11.紀翔瀛（2023年02月17日）。2022年第四季及全年我國扣件產業回顧與展望。ITIS智網。
https://www.itis.org.tw/NetReport/NetReport_Detail.aspx?rpno=356815962&type=netreport
12.徐佩吟（2021）。台灣扣件產業國際競爭力分析〔未出版之碩士論文〕。國立屏東大學國際貿易學系。
13.高宏宇（2023）。多模態語言強化模型之訊息科技前瞻技術研究（I）〔未出版之碩士論文〕。國立成功大學資訊工程學系。
14.陳宇勛（2020）。建構季節性產品之銷售預測模式：以18天生啤酒為例〔未出版之碩士論文〕。國立交通大學管理學院運輸物流學系。
15.馮世達（2007）。台灣螺絲螺帽產業競爭力分析〔未出版之碩士論文〕國立成功大學資源工程學系。
16.廖邵瑋（2022）。自然語言處理技術應用於科技華語詞彙分析〔未出版之碩士論文〕國立臺灣師範大學華語文教學系。
17.遠見雜誌（2021）。小螺絲螞蟻雄兵衝上全球第三大。遠見雜誌出版。https://event.gvm.com.tw/anniversary/35/feature/hardware.html
18.劉和財（2010）。台灣螺絲扣件產業成功關鍵因素之探討〔未出版之碩士論文〕國立高雄應用科技大學高階經營管理研究所。
19.劉麗惠（2017）。從「螺絲王國」到「醫療矽谷」台灣扣件產業轉型看見傳產新力量。貿易雜誌電子報，307，48-51。
http://www.ieatpe.org.tw/magazine/ebook307/b3.pdf
20.蔡政融（2021）。基於深度學習+BERT與強化學習進行FAANG股價預測。〔未出版之碩士論文〕國立政治大學資訊管理學系。
21.賴敏軒（2011）。實證探究多種鑑別式語言模型於語音辨識之研究〔未出版之碩士論文〕國立臺灣師範大學資訊工程研究所。
22.竇忠先（2007）。以策略管理觀點設計之銷售預測輔助系統-營業單位之銷售目標訂定。
23.黨建武、從筱卿（2021）。基於CNN 和GRU 的混合股指預測模型研究。電腦工程與應用，57(16)。
24.Huateng, O(2023)。AI預測力量:未來的商業與科學洞察:如何利用人工智能技術驅動資料分析與預測。Huateng Ou Press。
25.Adams, S., & Beling, P. A. (2019). A survey of feature selection methods for Gaussian mixture models and hidden Markov models. Artificial Intelligence Review, 52, 1739-1779.
26.Ashrafzadeh, M., Taheri, H. M., Gharehgozlou, M., & Zolfani, S. H. (2023). Clustering-based return prediction model for stock pre-selection in portfolio optimization using PSO-CNN+ MVF. Journal of King Saud University - Computer and Information Sciences, 35(9), 101737.
27.Bangi, M. S. F., & Kwon, J. S. I. (2023). Deep hybrid model-based predictive control with guarantees on domain of applicability. AIChE Journal, 69(5), e18012.
28.Bannour, N., Ghannay, S., Névéol, A., & Ligozat, A.L. (2021). Evaluating the carbon footprint of NLP methods: A survey and analysis of existing tools. In Proceedings of the Second Workshop on Simple and Efficient Natural Language Processing, 11-21.
29.Chen, Y., Wu, J., & Wu, Z. (2022). China’s commercial bank stock price prediction using a novel K-means-LSTM hybrid approach. Expert Systems with Applications, 202(C), 117370.
30.Dehghani, A., Moazam, H. M. Z. H., Mortazavizadeh, F., Ranjbar, V., Mirzaei, M., Mortezavi, S., Ng, J. L., & Dehghani, A. (2023). Comparative evaluation of LSTM, CNN, and ConvLSTM for hourly short-term streamflow forecasting using deep learning approaches. Ecological Informatics, 75, 102119.
31.DiPietro, R., & Hager, G. D. (2020). Deep learning: RNNs and LSTM. In Handbook of medical image computing and computer assisted intervention, 503-519 Elsevier.
32.Duchessi, P., & Lauría, E. J. (2013). Decision tree models for profiling ski resorts’ promotional and advertising strategies and the impact on sales. Expert Systems with Applications, 40(15), 5822-5829.
33.Elsaraiti, M., Ali, G., Musbah, H., Merabet, A., & Little, T. (2021). Time series analysis of electricity consumption forecasting using ARIMA model. In 2021 IEEE Green Technologies Conference, 259-262. IEEE.
34.Fan, W. (2022). Prediction of monetary fund based on ARIMA model. Procedia computer science, 208, 277-285.
35.Fang, W., Chen, Y., & Xue, Q. (2021). Survey on research of RNN-based spatio-temporal sequence prediction algorithms. Journal on Big Data,3(3), 97-110.
36.GeeksforGeeks. (2024, April 30). Top 20 LLM (Large Language Model) Models. GeeksforGeeks.
37.Gamboa, J. C. B. (2017). Deep learning for time-series analysis. arXiv preprint arXiv:1701.01887.
38.Gratton, L. (2021). How to do hybrid right. Harvard Business Review, 99(3), 65-74.
39.Hansun, S. (2013). A new approach of moving average method in time series analysis. 2013 Conference on New Media Studies (CoNMedia), 1-4.
40.Hoefler, R., Tiguman, G. M. B., Galvão, T. F., Ribeiro-Vaz, I., & Silva, M. T. (2023). Trends in sales of antidepressants in Brazil from 2014 to 2020: A time trend analysis with joinpoint regression. Journal of Affective Disorders, 323, 213-218.
41.Hoffmann, J., Borgeaud, S., Mensch, A., Buchatskaya, E., Cai, T., Rutherford, E., de Las Casas, D., Hendricks, L. A., Welbl, J., & Clark, A. (2022). An empirical analysis of compute-optimal large language model training. Advances in Neural Information Processing Systems, 35, 30016-30030.
42.Joseph, R. V., Mohanty, A., Tyagi, S., Mishra, S., Satapathy, S. K., & Mohanty, S. N. (2022). A hybrid deep learning framework with CNN and Bi-directional LSTM for store item demand forecasting. Computers & Electrical Engineering, 103(C), 14.
43.Kobiela, D., Krefta, D., Król, W., & Weichbroth, P. (2022). ARIMA vs LSTM on NASDAQ stock exchange data. Procedia Computer Science, 207, 3836-3845.
44.Kolkova, A. (2020). The Application of Forecasting Sales of Services to Increase Business Competitiveness. Journal of Competitiveness, 12(2), 90-105.
45.Liddy, E. D. (2001). Natural language processing. In Encyclopedia of Library and Information Science.
46.Lin, J., Nogueira, R., & Yates, A. (2022). Pretrained transformers for text ranking: Bert and beyond. Springer Nature.
47.Lin, Y. F., Cheng, C. S., & Chen, Y. C. (2021, September). Sales forecasting using ANNs or RNNs-A case study of freeway service station in Taiwan. In 2021 IEEE International Conference on Consumer Electronics-Taiwan (ICCE-TW), 1-2. IEEE.
48.Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., & Stoyanov, V. (2019). Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692.
49.Luna, J. C. (2023). 8 Top Open-Source LLMs for 2024 and Their Uses.
https://www.datacamp.com/blog/top-open-source-llms
50.Moghar, A., & Hamiche, M. (2020). Stock market prediction using LSTM recurrent neural network. Procedia computer science, 170, 1168-1173.
51.Muthukumar, P., & Zhong, J. (2021). A stochastic time series model for predicting financial trends using nlp. arXiv preprint arXiv:2102.01290.
52.Nirmala, V. W., Harjadi, D., & Awaluddin, R. (2021). Sales forecasting by using exponential smoothing method and trend method to optimize product sales in pt. zamrud bumi indonesia during the covid-19 pandemic. International Journal of Engineering, Science and Information Technology, 1(4), 59-64.
53.Palkar, A., Deshpande, M., Kalekar, S., & Jaswal, S. (2020). Demand forecasting in retail industry for liquor consumption using LSTM. 2020 International Conference on Electronics and Sustainable Communication Systems (ICESC), 521-525.
54.Panggabean, S., Sihombing, P. R., Dewi, K. H. S., Pramartha, I. N. B., Junaidy, J., & Syaharuddin, S.(2021). Simulasi Exponential Moving Avarage dan Single Exponential Smoothing: Sebuah Perbandingan Akurasi Metode Peramalan. Jurnal Pemikiran dan Penelitian Pendidikan Matematika (JP3M), 4(1), 1-10.
55.Sharma, M., Sharma, D., Burle, R., Patil, P., Joge, I., & Puri, C. (2024). Predicting house price model: A comprehensive analysis with random forest and decision tree method. 2024 3rd International Conference for Innovation in Technology (INOCON), 1-6.
56.Shih, S.-Y., Sun, F.-K., & Lee, H.-y. (2019). Temporal pattern attention for multivariate time series forecasting. Machine Learning, 108, 1421-1441.
57.Shumway, R. H., Stoffer, D. S., Shumway, R. H., & Stoffer, D. S. (2017). ARIMA models. Time series analysis and its applications: with R examples, 75-163.
58.Slater, L. J., Arnal, L., Boucher, M.-A., Chang, A. Y.-Y., Moulds, S., Murphy, C., Nearing, G., Shalev, G., Shen, C., & Speight, L. (2023). Hybrid forecasting: blending climate predictions with AI models. Hydrology and earth system sciences, 27(9), 1865-1889.
59.Suryana, N., & Basari, A. (2020). Generate contextual insight of product review using deep LSTM and word embedding. Journal of Physics: Conference Series, 1577(1), 012006. IOP Publishing.
60.Swari, M., Qusyairi, M., Mandyartha, E., & Wahanani, H. (2021). Business intelligence system using simple moving average method (Case study: Sales medical equipment at PT. Semangat Sejahtera Bersama). Journal of Physics: Conference Series, 1899(1), 012121. IOP Publishing.
61.Ullah, K., & Qasim, M. (2020). Google stock prices prediction using deep learning. 2020 IEEE 10th International Conference on System Engineering and Technology (ICSET), 108-113. IEEE.
62.Wang, X., & Zhang, Z. (2021, June 16). Must-Read Papers on Pre-Trained Language Models (PLMs). Github. https://github.com/thunlp/PLMpapers?tab=readme-ov-file.
63.Xie, H. H., Li, C., Ding, N., & Gong, C. (2021). Walmart sale forecasting model based on LSTM and LightGBM. 2021 2nd International Conference on Education, Knowledge and Information Management (ICEKIM), 366-369. IEEE.
64.Yang, S., Navarathna, P., Ghosh, S., & Bequette, B. W. (2020). Hybrid modeling in the era of smart manufacturing. Computers & Chemical Engineering, 140, 106874.
65.Yang, Y., Fan, C., & Xiong, H. (2022). A novel general-purpose hybrid model for time series forecasting. Applied Intelligence, 52(2), 2212-2223.
66.Zaremba, A., & Demir, E. (2023). ChatGPT: Unlocking the future of NLP in finance. Modern Finance, 1(1), 93-98.
67.Zhong, H., Xiao, C., Tu, C., Zhang, T., Liu, Z., & Sun, M. (2020). How does NLP benefit legal system: A summary of legal artificial intelligence. arXiv preprint arXiv:2004.12158.
68.Zhu, H. (2021). A deep learning based hybrid model for sales prediction of e-commerce with sentiment analysis. 2021 2nd International Conference on Computing and Data Science (CDS), 493-497. IEEE.