面向大宗商品价格的异常信息检测仿真

Simulation for Outlier Detection of Commodities Trading Price

大宗商品价格的异常波动会给相关企业和投资者带来较大经营与投资风险。大宗商品价格的波动状态具有非线性、高噪声、随机性等特点,价格所反映的异常信息具有局域累积性。目前,对大宗商品价格的预测是一个研究热点,而事实上,价格从长期来看可预测性较差,但是价格的异常波动在统计上是有一定规律的,且可以通过机器学习方法进行异常信息挖掘、仿真或预测。这方面的研究文献较少,也存在一定研究难度。所以,作者提出了一种基于小波分析与神经网络的大宗商品价格异常信息挖掘算法。首先,基于大宗商品价格数据的数理统计规律,利用小波变换的思想,分析数据的离散度。然后,生成神经网络输入矢量,构建训练集和测试集来训练神经网络模型。最后,利用训练好的预测模型预测大宗商品的异常状态。测试结果表明,上述模型具有一定的预测准确率,可以为企业和投资者的决策提供一定的参考。

Abnormal fluctuations of commodity price will bring some business risks and investment risks to the related enterprises or investors.The fluctuation state of commodity price has the characteristics of nonlinearity,high noise and randomness,and the abnormal information reflected by the price has the characteristic of local accumulation.At present,the prediction of commodity prices is a hot research topic,and in fact,the price is less predictable in the long term,but the abnormal fluctuation of price has certain regularity in statistics and it can be mined,simulated or predicted by machine learning methods.There are few studies in this field,and there are some difficulties.Therefore,an outlier mining method for bulk commodity price was put forward based on wavelet analysis and neural network.First,based on the mathematical statistics rule of commodity price data,the method used the idea of wavelet transform to analyze the deviation of data.Then input vectors were generated,and the training set and the test set were constructed to train the neural network model.Finally,the abnormal state of bulk commodities was predicted with the trained prediction model.The test results show that the model has satisfactory prediction accuracy and can provide some reference for enterprises and investors.