Research Of Web Log Mining Based On ACO

Internet has become an important platform for people to release, receive information and communication. According to statistics,99% of the information are useless to relatively 99% of the users, therefore, helping users to find the information they want to search or interested in from the Internet has becoming a hot research point in the field. A major challenge that various types of e-commerce sites face is the need to find out customers'interests and hobbies, discover users'accessing mode, and design personalized sites which satisfy the needs of different customers'groups, and this is the objective which every commercial site pursue. The method of web log mining can help to achieve this goal.Under the above background, the thesis studied web log mining based on ACO (ant colony Optimization algorithm), the main works are as follow:(1)By using the similarity between users'browsing behavior and ant foraging behavior, the thesis brings forward a new concept of "interest pheromone" to reflect users'browsing interest degree, by the use of browsing interest degree and select preference degree, this thesis designed a group users'browsing path algorithm based on ant colony algorithm. Experimental results show that the algorithm is feasible, using interest pheromones can accurately reflect users'browsing mode.(2)On the basis of basic ant colony clustering model, it improved the ant colony clustering algorithm from three aspects: characteristics of web users, similarity between objects, probability conversion function.(3)Incremental web user clustering algorithm was brought forward on the basis of an improved ant colony clustering algorithm (ACCA), it defines the user clustering center,at the same time introduces clustering disintegration mechanism and clustering models maintenance library to ensure that the incremental ant colony clustering algorithm can get high-quality clustering results. Experimental results show that the incremental ant colony clustering algorithm proposed by the thesis overcomes the deficiency of lack of scalability of the original ant colony clustering algorithm in handling large amounts of data. At the same time, by using the clustering model maintenance library, it makes the incremental user clustering algorithm get high-quality clustering results in a short period of time.