论文部分内容阅读
目的建立以临床和流行病学指标为基本分析因子的综合诊断及预测炭疽危害程度的智能预测模型,提高对炭疽病发生的认识和判断能力。方法根据实际疾病案例资料,分析临床症状、实验室检测指标、流行病学特征等因素。选入明显影响炭疽诊断和流行强度的指标,并将其作为神经元单位。利用Matlab 6．1软件中的神经网络工具箱训练、调整和建立智能化分析系统。结果多因素相关分析显示,疾病潜伏期、胸部X线检验结果、镜检结果、职业特征等11项指标与炭疽病的诊断和流行强度有关;神经网络经500步学习和训练,训练误差从6．669 59下降至5．05119×10-11,通过建立的智能神经网络模型对炭疽和非炭疽实际案例进行诊断和预测分析,其平均符合率达到100％。结论人工神经网络在疾病综合特征与炭疽诊断和危害度预测之间建模是可行的,所训练的智能模型预测平均符合率达100％,有很好的实际应用价值。
OBJECTIVE: To establish an intelligent prediction model based on comprehensive analysis of clinical and epidemiological indicators as the basic analysis factor and prediction of the extent of anthracnose hazards and to improve the understanding and judgment of anthrax occurrence. Methods According to the actual case data, analyze the clinical symptoms, laboratory test indicators, epidemiological characteristics and other factors. Targets that significantly affect the diagnosis and prevalence of anthrax were selected and used as neuronal units. Use Matlab 6.1 software in the neural network toolbox training, adjustment and establishment of intelligent analysis system. Results Multivariate correlation analysis showed that 11 indicators such as disease latency, chest X-ray test results, microscopic examination results and occupational characteristics were related to the diagnosis and prevalence of anthrax. The neural network was trained and trained after 500 steps with a training error of 6. 669 59 to 5.05119 × 10-11. The diagnostic and prognostic analysis of anthrax and non-anthrax cases were done by using the established intelligent neural network model. The average coincidence rate reached 100%. Conclusion Artificial neural network is feasible to model the comprehensive diagnosis of disease and anthrax diagnosis and prediction of the degree of harm. The trained intelligent model predicts the average coincidence rate of 100%, which has a good practical value.