Association Rule Discovery Application Model in E-Commerce Data Mining

This paper focuses on innovative association rule algorithms, integrating improved matrix decomposition with deep neural networks, aiming to efficiently mine potential high-order associations in e-commerce data. Through experimental simulation, compared with classic algorithms such as Apriori and FP-Growth, it shows significant advantages. In terms of support, the innovative algorithm performs well under low support thresholds. For example, when the support threshold is 0.01, the innovative algorithm discovers 1,200 association rules, far exceeding the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{3 5 0}$</tex> of the Apriori algorithm and the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{4 2 0}$</tex> of the FP-Growth algorithms, and can effectively mine low-frequency and valuable rules. In terms of confidence, taking the support of 0.05 as an example, the average confidence of the innovative algorithm reaches 0.68, and the proportion of rules in the high confidence interval (0.6-1) is significantly higher than that of traditional algorithms. In the comprehensive evaluation of F1 value, in the tests of multiple e-commerce transaction data sets, the innovative algorithm is significantly ahead. For example, the F1 value of the innovative algorithm on data set 1 is 0.75, while the Apriori algorithm is 0.52 and the FP-Growth algorithm is 0.55. Stability analysis shows that under different data set sizes and time spans, the support, confidence and F1 value of the innovative algorithm fluctuate very little, showing strong stability. This study provides e-commerce companies with more reliable and efficient data mining tools to help them make accurate decisions in complex market environments.