根據統計，職業傷害每年造成全台灣五萬多勞工輕重傷或死亡與數十萬總損失工作日數。台灣所有產業中又以製造業所造成之職業傷害所佔比例最高。鋼鐵工業屬於製造業之一部份，並且其失能傷害頻率與失能傷害嚴重度都較一般產業嚴重許多，其員工無疑處於一個極度容易發生危險的環境。對於意外發生之相關應對措施，相較於等待意外發生後進行檢討與改善，如果能在意外發生之前就採取行動預防會是更好的選擇。因此，意外發生的預測逐漸成為關鍵的議題之一。安全氣候為一評估組織安全之方法，並且有許多研究得到安全氣候與受傷率有一定程度之相關性。對於預測意外之相關議題，學者常使用不同之迴歸分析方法來處理。近年來類神經網路出現頻繁，由於其可以透過學習自行建立輸入與目標之間之關係，類神經網路也常被應用於預測。因此本研究欲透過迴歸分析與類神經網路以不同的安全氣候因子預測鋼鐵產業之第一線作業員是否會發生意外，並根據結果比較何者較適合應用於本研究。結果顯示，根據不同的預測因子，迴歸分析的準確率由高至低依序為以全部分數、自身相關分數、組織氣候分數、團體氣候分數、全部分數平均做為預測因子，敏感度的結果則與準確率相反，類神經網路的準確率由高至低則依序為以全部分數、自身相關分數、團體氣候分數、組織氣候分數、全部分數平均做為預測因子，敏感度依序為以全部分數平均、全部分數、自身相關分數、團體氣候分數、組織氣候分數做為預測因子，將兩種方法的準確率與敏感度綜合討論後，本研究認為透過類神經網路以所有問卷問題分數做為預測因子以預測員工之意外發生與否有最佳之表現(準確率70.96%、敏感度60%)。

According to the statistics reported by Taiwan Ministry of Labor, the occupational accidents have caused over 50,000 workers minor, serious or even fatal injuries every year. These accidents result in a loss of hundreds of thousands of working days per year. In Taiwan, the manufacturing industry has the highest proportion of occupational accidents of all industries. Steel industry is one of the manufacturing industry and the disabling injury frequency rate and severity rate of steel industry are higher than the average of all industries. While statistical methods are used to analyze accidents and improve safety, it is best to prevent accidents before they happen. Accident prediction is very important to helping prevent accidents in these industries.
Safety climate is used to evaluate the safety of the organization. Research shows a significant correlation between safety climate and accident rates. Different regression models have been used for the prediction and analysis of accidents. In addition, artificial neural networks that are often used in predictions can identify the relationship between input and target data. The purpose of this study is to use regression models and artificial neural network models to predict if steel industry workers will encounter accidents. Different predictors were generated using a safety climate questionnaire, and the results were compared to determine the most suitable method.
The sensitivity results of the regression model were the opposite of the accuracy results of the regression model which was ranked from highest to lowest using the different predictors as: the total questionnaire scores, employee safety score, organizational-level safety climate score, group-level safety climate score, and the average of all questionnaire scores. The accuracy results of the artificial neural network model from highest to lowest were: all scores from the questionnaire, employee safety score, group-level safety climate score, organizational-level safety climate score, and the average of all questionnaire scores. The sensitivity results of the artificial neural network model, from highest to lowest, used the average of all questionnaire scores, all scores from the questionnaire, employee safety scores, group-level safety climate scores, and organizational-level safety climate scores. After a comprehensive discussion of accuracy results and sensitivity results, it was determined that using artificial neural network model based on all the questions of the safety climate questionnaire as predictor was the most suitable method of the study.

Abstract ii
圖目錄 vi
表目錄 vii
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 6
1.3 研究範圍與限制 6
1.4 研究流程 7
第二章 文獻探討 9
2.1 安全氣候 9
2.2 迴歸分析 11
2.3 類神經網路 13
第三章 研究方法 17
3.1 安全氣候問卷 17
3.2 模型建立 18
3.2.1 迴歸分析 19
3.2.2 類神經網路 19
3.3 模型驗證 22
3.4 分析預測結果 23
第四章 結果 24
4.1 受試者背景資料 24
4.2 迴歸分析預測結果 26
4.2.1 以組織層級安全氣候分數預測意外發生 26
4.2.2 以團體層級安全氣候分數預測意外發生 27
4.2.3 以員工自身相關之安全分數預測意外發生 27
4.2.4 以所有問卷問題分數測意外發生 28
4.2.5 以各個安全分數平均後預測意外發生 28
4.3類神經網路預測結果 29
4.3.1 以組織層級安全氣候分數預測意外發生 29
4.3.2 以團體層級安全氣候分數預測意外發生 33
4.3.3 以員工自身相關之安全分數預測意外發生 37
4.3.4 以所有問卷問題分數測意外發生 41
4.3.5 以各個安全分數平均預測意外發生 45
4.4多重線性迴歸與類神經網路之結果比較 48
4.4.1 整體準確率 48
4.4.2 敏感度 50
第五章 討論 52
5.1 迴歸分析之結果 52
5.2 類神經網路之結果 53
5.3 預測因子 55
第六章 結論 58
6.1 研究結論 58
6.2 後續研究方向 59
參考文獻 62
附錄一 安全氣候問卷 70

1. 張斐章、張麗秋(2015)，類神經網路導論-原理與應用，第二版
2. 陳玉伶(2016)，鋼鐵業安全氣候與工安意外之關係研究
3. 勞動部職業安全衛生署(2012)，中華民國100年勞動檢查年報
4. 勞動部職業安全衛生署(2013)，中華民國101年勞動檢查年報
5. 勞動部職業安全衛生署(2014)，中華民國102年勞動檢查年報
6. 勞動部職業安全衛生署(2014)，鋼鐵業安全衛生危害風險評估與控制技術手冊
7. 勞動部職業安全衛生署(2015)，中華民國103年勞動檢查年報
8. 勞動部職業安全衛生署(2016)，中華民國104年勞動檢查年報
9. 顏秀珍、李御璽、王秋光(2009)，改善不平衡資料集中少數類別資料之分類正確性的方法。電子商務學報，第十一卷，第四期
10. Al-Ghamdi, A. S. (2002). Using logistic regression to estimate the influence of accident factors on accident severity. Accident Analysis & Prevention, 34(6), 729-741.
11. Anderson, H. R., de Leon, A. P., Bland, J. M., Bower, J. S., & Strachan, D. P. (1996). Air pollution and daily mortality in London: 1987-92. Bmj, 312(7032), 665-669.
12. Arcury, T. A., O’Hara, H., Grzywacz, J. G., Isom, S., Chen, H., & Quandt, S. A. (2012). Work safety climate, musculoskeletal discomfort, working while injured, and depression among migrant farmworkers in North Carolina. American journal of public health, 102(S2), S272-S278.
13. Baxt, W. G. (1991). Use of an artificial neural network for the diagnosis of myocardial infarction. Annals of internal medicine, 115(11), 843-848.
14. Baxt, W. G., & Skora, J. (1996). Prospective validation of artificial neural network trained to identify acute myocardial infarction. The Lancet, 347(8993), 12-15.
15. Bottaci, L., Drew, P. J., Hartley, J. E., Hadfield, M. B., Farouk, R., Lee, P. W., ... & Monson, J. R. (1997). Artificial neural networks applied to outcome prediction for colorectal cancer patients in separate institutions. The Lancet, 350(9076), 469-472.
16. Brislin, R. W. (1970). Back-translation for cross-cultural research. Journal of cross-cultural psychology, 1(3), 185-216.
17. Brown, R. L., & Holmes, H. (1986). The use of a factor-analytic procedure for assessing the validity of an employee safety climate model. Accident Analysis & Prevention, 18(6), 455-470.
18. Bryson, A. E., & Ho, Y. C. (1969). Applied Optimal Control, Optimization, Estimation and Control, Waltham.
19. Cabrera, D. D., Isla, R., & Vilela, L. D. (1997). An evaluation of safety climate in ground handling activities. Aviation safety, 255-268.
20. Chang, L. Y. (2005). Analysis of freeway accident frequencies: negative binomial regression versus artificial neural network. Safety science, 43(8), 541-557.
21. 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.
22. Chen, H., Cao, L., & Logan, D. B. (2012). Analysis of risk factors affecting the severity of intersection crashes by logistic regression. Traffic injury prevention, 13(3), 300-307.
23. Chester, D. L. (1990, January). Why two hidden layers are better than one. In Proceedings of the international joint conference on neural networks (Vol. 1, pp. 265-268).
24. Christian, M. S., Bradley, J. C., Wallace, J. C., & Burke, M. J. (2009). Workplace safety: a meta-analysis of the roles of person and situation factors. Journal of Applied Psychology, 94(5), 1103.
25. Collaborators, M. C. T., Perel, P., Arango, M., Clayton, T., Edwards, P., Komolafe, E., ... & Yutthakasemsunt, S. (2008). Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. Bmj, 336(7641), 425-9.
26. Cooper, M. D., & Phillips, R. A. (1994, January). Validation of a safety climate measure. In Occupational Psychology Conference of the British Psychological Society (Vol. 3, No. 5).
27. Coyle, I. R., Sleeman, S. D., & Adams, N. (1996). Safety climate. Journal of Safety Research, 26(4), 247-254.
28. Davies, R. J., Ali, N. J., & Stradling, J. R. (1992). Neck circumference and other clinical features in the diagnosis of the obstructive sleep apnoea syndrome. Thorax, 47(2), 101-105.
29. Dedobbeleer, N., & Béland, F. (1991). A safety climate measure for construction sites. Journal of safety research, 22(2), 97-103.
30. Diller, T., Helmrich, G., Dunning, S., Cox, S., Buchanan, A., & Shappell, S. (2013). The human factors analysis classification system (HFACS) applied to health care. American Journal of Medical Quality, 29(3), 181-190.
31. Esfe, M. H., Rostamian, H., Toghraie, D., & Yan, W. M. (2016). Using artificial neural network to predict thermal conductivity of ethylene glycol with alumina nanoparticle. Journal of Thermal Analysis and Calorimetry, 126(2), 643-648.
32. Fang, D., Chen, Y., & Wong, L. (2006). Safety climate in construction industry: a case study in Hong Kong. Journal of construction engineering and management, 132(6), 573-584.
33. Gauss, C. F. (1809). Theoria motus corporum coelestium in sectionibus conicis solem ambientium.
34. Gholipour, A., & Arjmand, N. (2016). Artificial neural networks to predict 3D spinal posture in reaching and lifting activities; Applications in biomechanical models. Journal of Biomechanics, 49(13), 2946-2952.
35. Glennon, D. P. (1982). Measuring organisational safety climate. Australian Safety News, 23, 23-28.
36. Hayashi, Y., Sakata, M., & Gallant, S. I. (1990). Multi-layer versus single-layer neural networks and an application to reading hand-stamped characters. In International Neural Network Conference (pp. 781-784).
37. Hebb, D. O. (1949). The organization of behavior: A neuropsychological theory.
38. Hecht-Nielsen, R. (1989, June). Theory of the backpropagation neural network. In International Joint Conference on Neural Networks (pp. 593-605)
39. Heinrichs, M., Wagner, D., Schoch, W., Soravia, L. M., Hellhammer, D. H., & Ehlert, U. (2005). Predicting posttraumatic stress symptoms from pretraumatic risk factors: a 2-year prospective follow-up study in firefighters. American Journal of Psychiatry, 162(12), 2276-2286.
40. Hopfield, J. J. (1982). Neural networks and physical systems with emergent collective computational abilities. Proceedings of the national academy of sciences, 79(8), 2554-2558.
41. Hopfield, J. J. (1984). Neurons with graded response have collective computational properties like those of two-state neurons. Proceedings of the national academy of sciences, 81(10), 3088-3092.
42. Hornik, K., Stinchcombe, M., & White, H. (1990). Universal approximation of an unknown mapping and its derivatives using multilayer feedforward networks. Neural networks, 3(5), 551-560.
43. Huang, Y. H., Chen, P. Y., Krauss, A. D., & Rogers, D. A. (2004). Quality of the execution of corporate safety policies and employee safety outcomes: Assessing the moderating role of supervisor safety support and the mediating role of employee safety control. Journal of Business and Psychology, 18(4), 483-506.
44. Huang, Y. H., Zohar, D., Robertson, M. M., Garabet, A., Lee, J., & Murphy, L. A. (2013). Development and validation of safety climate scales for lone workers using truck drivers as exemplar. Transportation research part F: traffic psychology and behaviour, 17, 5-19.
45. Hush, D. R., & Horne, B. G. (1993). Progress in supervised neural networks. IEEE signal processing magazine, 10(1), 8-39.
46. Imprialou, M. I. M., Quddus, M., & Pitfield, D. E. (2016). Predicting the safety impact of a speed limit increase using condition-based multivariate Poisson lognormal regression. Transportation Planning and Technology, 39(1), 3-23.
47. Jimmieson, N. L., Tucker, M. K., White, K. M., Liao, J., Campbell, M., Brain, D., Page, K., Barnett, A.G. & Graves, N. (2016). The role of time pressure and different psychological safety climate referents in the prediction of nurses’ hand hygiene compliance. Safety Science, 82, 29-43.
48. Jovanis, P. P., & Chang, H. L. (1986). Modeling the relationship of accidents to miles traveled. Transportation Research Record, 1068, 42-51.
49. Kalogirou, S. A., & Bojic, M. (2000). Artificial neural networks for the prediction of the energy consumption of a passive solar building. Energy, 25(5), 479-491.
50. Kang, S. Y. (1992). An investigation of the use of feedforward neural networks for forecasting.
51. Kimoto, T., Asakawa, K., Yoda, M., & Takeoka, M. (1990, June). Stock market prediction system with modular neural networks. In International Joint Conference on Neural Networks (pp. 1-6).
52. Kůrková, V. (1992). Kolmogorov's theorem and multilayer neural networks. Neural networks, 5(3), 501-506.
53. Law, R., & Au, N. (1999). A neural network model to forecast Japanese demand for travel to Hong Kong. Tourism Management, 20(1), 89-97.
54. Legendre, A. M. (1805). Nouvelles méthodes pour la détermination des orbites des comètes.
55. Lenné, M. G., Salmon, P. M., Liu, C. C., & Trotter, M. (2012). A systems approach to accident causation in mining: An application of the HFACS method. Accident Analysis & Prevention, 48, 111-117.
56. Lippmann, R. (1987). An introduction to computing with neural nets. IEEE Assp magazine, 4(2), 4-22.
57. McClelland, J. L., & Rumelhart, D. E. (1986). Parallel distributed processing: explorations in the microstructure of cognition.
58. McCulloch, W. S., & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous activity. The bulletin of mathematical biophysics, 5(4), 115-133.
59. Miaou, S. P. (1994). The relationship between truck accidents and geometric design of road sections: Poisson versus negative binomial regressions. Accident Analysis & Prevention, 26(4), 471-482.
60. Miaou, S. P., & Lum, H. (1993). Modeling vehicle accidents and highway geometric design relationships. Accident Analysis & Prevention, 25(6), 689-709.
61. Minsky, M., & Papert, S. (1969). Perceptrons.
62. Neal, A., & Griffin, M. A. (2006). A study of the lagged relationships among safety climate, safety motivation, safety behavior, and accidents at the individual and group levels. Journal of applied psychology, 91(4), 946.
63. Neal, A., Griffin, M. A., & Hart, P. M. (2000). The impact of organizational climate on safety climate and individual behavior. Safety science, 34(1), 99-109.
64. Niskanen, T. (1994). Safety climate in the road administration. Safety Science, 17(4), 237-255.
65. Palei, S. K., & Das, S. K. (2009). Logistic regression model for prediction of roof fall risks in bord and pillar workings in coal mines: An approach. Safety Science, 47(1), 88-96.
66. Patel, D. A., & Jha, K. N. (2016). Evaluation of construction projects based on the safe work behavior of co-employees through a neural network model. Safety science, 89, 240-248.
67. Rosenblatt, F. (1958). The perceptron: a probabilistic model for information storage and organization in the brain. Psychological review, 65(6), 386.
68. Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1988). Learning representations by back-propagating errors. Cognitive modeling, 5(3), 1.
69. Schöneburg, E. (1990). Stock price prediction using neural networks: A project report. Neurocomputing, 2(1), 17-27.
70. Tang, Z., & Fishwick, P. A. (1993). Feedforward neural nets as models for time series forecasting. ORSA journal on computing, 5(4), 374-385.
71. Vredenburgh, A. G. (2002). Organizational safety: which management practices are most effective in reducing employee injury rates?. Journal of safety Research, 33(2), 259-276.
72. Williams, D. R. G. H. R., & Hinton, G. E. (1986). Learning representations by back-propagating errors. Nature, 323, 533-536.
73. Williamson, A. M., Feyer, A. M., Cairns, D., & Biancotti, D. (1997). The development of a measure of safety climate: the role of safety perceptions and attitudes. Safety Science, 25(1), 15-27.
74. Wong, F. S. (1991). Time series forecasting using backpropagation neural networks. Neurocomputing, 2(4), 147-159.
75. Zegeer, C.V., Stewart, R., Reinfurt, D., Council, F., Neuman, T., Hamilton, E., Miller, T., Hunter, W., (1990). Cost effective geometric improvements for safety upgrading of horizontal curves.
76. Zhang, G., & Hu, M. Y. (1998). Neural network forecasting of the British pound/US dollar exchange rate. Omega, 26(4), 495-506.
77. Zohar, D. (1980). Safety climate in industrial organizations: theoretical and applied implications. Journal of applied psychology, 65(1), 96.
78. Zohar, D. (2008). Safety climate and beyond: A multi-level multi-climate framework. Safety Science, 46(3), 376-387.
79. Zohar, D., & Luria, G. (2005). A multilevel model of safety climate: cross-level relationships between organization and group-level climates. Journal of Applied Psychology, 90(4), 616.