Research On Some Key Technology Of Data Mining In E-commerce

At present, the practical application has proved, data mining in e-commerce and other aspects has very high commercial value and significance, and has been widely applied to various fields. For data mining, its algorithms are endless, especially such as association rule mining algorithm. Commonly, classical mining algorithm is Apriori algorithm in mining association rules. For Apriori algorithm, its key is how to obtain frequent item sets based on the minimum support, and currently, the minimum support of these algorithm was artificially set, according to users’mining experience. This will certainly give those results relatively large deviation. In some cases, the deviation is even fatal. Therefore, how to make its minimal impact on the minimum support for artificial setting? Thus, for this problems, the paper launched a more detailed discussion, and be addressed through the full use of the data mining technology and other methods, its concerned work and results are as follows:(1) Algorithm’s design mode. To effectively improve the scalability and maintainability of algorithm design, to avoid algorithm design’s confusion, to reduce the deviation, by exploring the combination of decorative pattern, strategic mode and adapter mode, this paper proposes three options, makes a detailed analysis and comparison of its feasibility and efficiency. Finally, we use a more efficient third schedule to apply to the development of the system, and this schedule has been verified by experiment.(2) Software complexity and AOP (Aspect Oriented Program). To control Powerfully the high level of complexity of the system, to maintain high cohesion and low coupling of modules, to avoid the high redundancy of the algorithm, to reduce the impact of other factors, according to the characteristics and needs of this case, this job discussed some important indicators of software complexity to achieve expected goals by AOP. By using the analytic hierarchy in economics, the part made the quantitative analysis for AOP playing an important role in decreasing the level of software complexity. Experiments show that, AOP has made system to achieve the desired effect. (3) Mining Association Rules’algorithm. During Mining association rules, the problem of the minimum support re-identify has been solved through the reverse self-learning solutions of the Newton interpolation algorithm. When Newton interpolation algorithm determines the point value, the entire transaction database needs being repeated for its scanning, which generates huge data. Therefore, the algorithm has vital impact on time complexity of association rule mining algorithm. So, this work is to solve the high time complexity’s problem through the use of the block algorithm. And the point values and results belong controllable range. Finally, it greatly reduces the time complexity of concerned algorithm. Experiments show that improvement measures are feasible and effective.(4) System development. To achieve the desired level and the above discussion results, an e-commerce system was developed. Through the comparison, analysis and mapping of the experimental procedure and experimental results’data, compared with the existing method, the various improvements are feasible and fruitful.At the end, this paper did some summaries.Some suggestions or ideas were proposed for some demerits, and good direction and motivation were given for the next seminar work.