Nowadays, personalized recommender technology based on Web log mining has been widely used in the e-commerce website. For the issues that the existing recommender systems do not have high accuracy, a recommendation system for e-commerce based on web log mining and correlation measure is proposed. First, the user's access log is extracted, and the data is preprocessed to obtain the structured data. Then, the log is analyzed to extract the characteristic sequence. After that, the correlation between the page and the transaction text documents is calculated according to the occurrence frequency of characteristics and the page dwell time. Finally, the angle cosine formula is used to calculate the correlation between the user and the page, and thus form a list of recommendations. Experimental results show that the proposed scheme can accurately give personalized recommendation according to the user's preference.