Abnormal behaviour detection of e-commerce consumers based on improved hidden Markov model

To address the issues of low anomaly detection rate, high false positive rate, and long detection time in traditional methods, an abnormal behaviour detection method for e-commerce consumers based on an improved hidden Markov model is proposed. The Scrapy spider framework is used to collect e-commerce consumer behaviour data, including purchase data, browsing data, search data, and evaluation data. The collected data is processed using an improved K-means algorithm for clustering, with normalisation, missing value imputation, and outlier removal applied to the clustering results. The MOPSO algorithm is used to optimise the parameters of the hidden Markov model, and the processed data is then input into the improved hidden Markov model to output the relevant detection results. Experimental results show that the maximum anomaly detection rate of this method is 96.7%, the maximum false positive rate is 4.7%, and the average detection time is 0.73 s.