臺灣博碩士論文加值系統

電器老舊或電器使用行為不當所造成火災或故障傷害，成為影響家庭安全的主要因素，如何有效偵測家電異常情況並預先警示使用者進行更換或改善，成為一項重要研究議題。本研究針對此議題結合物聯網與大數據分析技術，應用智慧電表資料分析提出了一套主動式學習的家電異常偵測方法，改善以往方法樣本收集不易的情況。以家庭常見的電器-電風扇為例進行相關驗證，其結果顯示我們的方法較傳統方法能有效改善偵測誤差。異常偵測的結果可讓使用者參考以進行相關電器保養及更換措施，避免因電器故障所引發的危害。

Fire and accidental damage caused by appliance aging or improper operating are the main factors of home security. Therefore, how to detect the anomaly of appliances and promptly warn the users to replace or pay attention to the improvement of the appliances become an important research topic. In this study, we combine the Internet of Things and big data analysis technology to this issue and apply smart meter data analysis to propose an anomaly detection method based on active learning to detect home appliance operation anomaly and to overcome the situation that collecting anomaly label is not easy. The experience use fans as measurement targets for proposed method. The results show that this approach compared to traditional anomaly detection method effectively improve the detection error. Users can refer to the results of anomaly detection to carry out the relevant electrical maintenance and replacement measures to avoid the electrical fault happens and cause harm.

誌謝 i
中文摘要 ii
ABSTRACT iii
CONTENTS iv
LIST OF FIGURES vii
LIST OF TABLES ix
Chapter 1 緒論 1
1.1 研究動機 1
1.2 目標與貢獻 2
1.3 論文架構 3
Chapter 2 文獻探討 4
2.1 異常偵測 4
2.1.1 異常型態 4
2.1.2 異常偵測方法 5
2.2 電力資訊異常偵測 7
2.3 主動式學習 7
2.4 主動式學習應用於異常偵測 9
Chapter 3 研究方法 11
3.1 系統流程 11
3.2 資料處理相關 12
3.2.1 資料選取與清理 12
3.2.2 資料轉換與特徵萃取 13
3.3 主動式學習異常偵測系統 15
3.3.1 一元分類模型 15
3.3.2 導入主動式學習 16
3.3.3 class imbalance問題 17
3.3.4 隨機森林搭配分層抽樣方法 19
3.3.5 演算法步驟 20
Chapter 4 實驗方法與結果 22
4.1 方法評估標準 22
4.1.1 Confusion matrix 22
4.1.2 ROC曲線以及AUC值 23
4.1.3 k-fold cross-validation 24
4.2 系統實驗 25
4.2.1 電表資料集測試 25
4.2.2 不同查詢樣本個數比較 27
4.2.3 滑動視窗大小對準確度影響 29
4.2.4 與非監督式方法比較 32
4.2.5 與監督式方法比較 33
4.2.6 其他資料集測試 34
4.2.7 實驗結果 42
Chapter 5 結論與未來發展 43
5.1 基於 pseudo label 的主動式學習電器異常偵測 43
5.2 未來發展 43
REFERENCE 44

[1]中華民國內政部消防署 (2016)。101-105年全國火災次數、起火原因及火災損失統計表。取自 http://www.nfa.gov.tw/main/List.aspx?ID=&MenuID=342
[2]Syarif, I., Prugel-Bennett, A., & Wills, G. (2012). Unsupervised clustering approach for network anomaly detection. Networked Digital Technologies, 135-145.
[3]Chandola, V., Banerjee, A., & Kumar, V. (2009). Anomaly detection: A survey. ACM computing surveys (CSUR), 41(3), 15.
[4]Goldstein, M., & Uchida, S. (2016). A comparative evaluation of unsupervised anomaly detection algorithms for multivariate data. PloS one, 11(4).
[5]Chou, J. S., & Telaga, A. S. (2014). Real-time detection of anomalous power consumption. Renewable and Sustainable Energy Reviews, 33, 400-411.
[6]Peña, M., Biscarri, F., Guerrero, J. I., Monedero, I., & León, C. (2016). Rule-based system to detect energy efficiency anomalies in smart buildings, a data mining approach. Expert Systems with Applications, 56, 242-255.
[7]Settles, B. (2010). Active learning literature survey. University of Wisconsin, Madison, 52(55-66), 11.
[8]Abe, N., Zadrozny, B., & Langford, J. (2006). Outlier detection by active learning. Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining - KDD 06, 504-509.
[9]Barnabe-Lortie, V., Bellinger, C., & Japkowicz, N. (2015). Active Learning for One-Class Classification. 2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA), 390-395.
[10]Schölkopf, B., & Smola, A. J. (2002). Learning with kernels: support vector machines, regularization, optimization, and beyond. MIT press.
[11]Yan, Y., Xu, Z., Tsang, I. W., Long, G., & Yang, Y. (2016, January). Robust Semi-Supervised Learning through Label Aggregation. AAAI, 2244-2250.
[12]Araya, D. B., Grolinger, K., Elyamany, H. F., Capretz, M. A., & Bitsuamlak, G. (2016). Collective contextual anomaly detection framework for smart buildings. 2016 International Joint Conference on Neural Networks (IJCNN), 511-518.
[13]王佑鈞、莊棨椉、張瑞益(2016)。低解析度電器特徵值評估及其在電器狀態辨識的應用。資訊與管理科學，9.2，52-63。
[14]Khan, S. S., & Madden, M. G. (2014). One-class classification: taxonomy of study and review of techniques. The Knowledge Engineering Review, 29(3), 345-374.
[15]Hawkins, S., He, H., Williams, G., & Baxter, R. (2002, September). Outlier detection using replicator neural networks. In DaWaK (Vol. 2454, pp. 170-180).
[16]He, H., & Garcia, E. A. (2009). Learning from imbalanced data. IEEE Transactions on knowledge and data engineering, 21(9), 1263-1284.
[17]Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16, 321-357.
[18]Wang, B. X., & Japkowicz, N. (2004, June). Imbalanced data set learning with synthetic samples. In Proc. IRIS Machine Learning Workshop (Vol. 19).
[19]Liu, X. Y., Wu, J., & Zhou, Z. H. (2009). Exploratory undersampling for class-imbalance learning. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 39(2), 539-550.
[20]Breiman, L. (2001). Random forests. Machine learning, 45(1), 5-32.
[21]Fawcett, T. (2016, August). Learning from Imbalanced Classes. Silicon Valley Data Science. Retrieved from https://svds.com/learning-imbalanced-classes/
[22]Arlot, S., & Celisse, A. (2010). A survey of cross-validation procedures for model selection. Statistics surveys, 4, 40-79.
[23]Liu, F. T., Ting, K. M., & Zhou, Z. (2008). Isolation Forest. 2008 Eighth IEEE International Conference on Data Mining, 413-422.
[24]Lai, L. B., Chang, R. I., & Kouh, J. S. (2008). Detecting network intrusions using signal processing with query-based sampling filter. EURASIP Journal on Advances in Signal Processing, 2009(1), 735283.
[25]UC Irvine Machine Learning Repository. (n.d.). Retrieved from http://archive.ics.uci.edu/ml/index.php