Formula Discovery Of Data Mining Algorithms And Improvement Of The Fdd.1

In the scientific history, each kind of the order of nature of physics, chemistry, astronomy are found by the scientists with researching a mass of data, such as the three laws of Newton, the law of universal gravitation, the planetary motion law of Kepler. These natural laws are the cornerstone of the science development and the society progresses.With the computer appearance, the data fitting technology has been developed, it is the important computation branch. With the technology of the data fitting, the approximate formula of the independent variable and a dependent variable can be found in the massive surveys data of the scientific experiment. This kind of formula is uniformly expressed as the algebra polynomial form, its coefficient is obtained with establishing regular system of equations using least square method. The algebra multinomial has a problem, that is when its time increases, the coefficient of the determinants of the system of linear equations will have "the morbid state" (namely small changes of determinant element will lead to a significant change in the solution).The empirical formula discovered theory FDD (Formula Discovery from Data) is the system which is based on machine discovered technology of artificial intelligence technology, curve fitting technology of the data computation and the visible technology. It discovers the empirical formula from the massive empirical datum, gradually completes the willfully combination of the arbitrary function, realizes the discovery of the natural law and the experience rule.This paper elaborated the curve fitting basic concept, the traditional fitting method, the application and the algorithm insufficiency and so on. The paper introduced BACON system and the FDD formula discovery system which are based on the artificial intelligence, analyzed FDD.1 formula discovery algorithm and other two editions FDD.2 and FDD.3 which are based on FDD.1.This paper, in view of to the FDD algorithm limitation, summarized the problems which exist in FDD. The purpose is enhancing the application scope of the algorithm. In order to solving the problem that existed in the infinite circulation study, I made the improvement to the heuristic search information of the algorithm. Again in this foundation I expanded the prototype storehouse, so the application scope of the algorithm is wider. I has carried on the realization onPC machine to the improvement FDD system, and has carried on the massive experiments.