Researches On Application Of Data Mining Technology In The Individual Network Teaching Platform

Based on mathematics, the optimization technology is used for solving various engineering problems, whose aim is to select the optimal one from the many solutions for specific problem. In early researches about the optimization problems, classical optimization algorithms(such as linear programming and dynamic programming) can get the optimal solution for solving small scale problems. However, with the size of the problems increasing gradually, the proficiency of classical optimization algorithms may be not satisfactory. Moreover, the classical optimization algorithms are always incapable to solve many practical problems, which are belonging to the multi-objective, nonlinear, and even more complex problems. Hence, finding suitable and intelligent algorithms have become important and attractive researches in some relevant subjects.Proposed in recent decades of years, intelligent optimization algorithms are novel optimization algorithms, which are also called meta-heuristic algorithms. These algorithms are developed through the simulations of the nature, and provide new ideas to solve complex problems. For example, these algorithms include simulated annealing algorithm, genetic algorithm, particle swarm algorithm and ant colony algorithm. Harmony search algorithm(HS), proposed by the South Korean scholar Geem[1] in 2001, is a new swarm intelligence optimization algorithm. It simulates the process of musicians adjusting tones of each instruments until they get a wonderful harmony. It has the advantages of less parameters, easy implementation and fast convergence speed, and successfully applied in many optimization problems.Based on extensive readings and analyzing the domestic and foreign papers researching for the HS algorithm, we analyze and summarize the basic principle, algorithm parameters and operation processes. For solving the two basic models of the continuous optimization problems and the multidimensional knapsack problems(MKP) belonging to the discrete optimization problems, the basic HS algorithm is adapted respectively. The main contents are summarized as follows.1、For solving continuous optimization problems, we propose a novel memory consideration strategy based on a elimination rule, local search strategy and tournament rule, in which the better harmony can have bigger probabilities to generate new harmonies. Then two important parameters bw and PAR are adjusted self-adaptively and dynamically, respectively. Finally, the comparative evaluations indicate the the proposed algorithm is much more effective.2、For solving the constrained optimization problems, we propose an improved penalty method, namely two-stage penalty method, which can not only exploit the positive information contained in the infeasible solutions, but also can remain the algorithm proficiency. Then, based on the feature of two-stage penalty, we develop an improved harmony memory consideration rule and adjust the parameters dynamically, which can enhance the global search ability in the first stage and the local search ability in the second stage. Finally, the extensive experiments indicate the effectiveness of the two-stage penalty method and the proficiency of the proposed algorithm.3、For solving the multidimensional knapsack problems(MKP), we present a novel harmony generating strategy, which abandon the mutation strategy for solving combinatorial optimization problem. It can enhance the performance of the algorithm effectively. Then, to balance the global search and local search abilities, the FFO strategy is integrated into the improved harmony search algorithm. Finally, the simulated experiments demonstrate the effectiveness and efficiency of the novel algorithm.