一种基于SVM的电力行业物资需求预测方法

摘要

为了减少物资需求审核工作量，提高审核效率和准确率，提出一种基于SVM的电力行业物资需求预测方法。该方法首先分析历史样本数据，把物资需求审核转换分类问题，然后对数据预处理，结合电力领域知识库，定义及提取需求特征，最后通过支持向量机训练出模型，实现对物资采购数量和种类的审核。实验结果表明，该方法审核精度为87.3%，说明利用领域知识库，基于能够SVM的电力行业物资需求预测方法能够有效提高审核效率和准确率。%The method, based on SVM, a kind of electric power industry material demand forecasts ,has been proposed, in order to reduce audit work of the material demand, improving the efficiency and accuracy. Firstly, the method analyzed historical sample data and translated materials demand audit into classification problem. Secondly, it need preprocessing the data, making it standardization. Defining and extracting demand characteristics by combining power domain knowledge base. Finally, support vector machine, by training model, finished the audit work on types and amounts of material purchase.