The Application Of Data Mining Technology In Reference Sky Classification

Due to the rapid expansion of the computer database in recent decades, people aredrowned in data of marine.They know their grasp of the data is far behind the speed ofdata growth. The amounts of data is increasing everyday and at the same time lots ofdata in the database are neglected and have not obtained effective utilization, whichlead to the loss of hidden information in these data. Therefore, the emergence of datamining complies with the current requirements of society.These requirements promotethe rapid development of the data mining technology.Today, the energy and environmental issues are becoming more and more serious.People are looking for a clean and efficient renewable energy to replace the limitedenergy on the earth. Solar energy, a kind of the inexhaustible renewable green energy,not only brings heat energy but also light energy for humans.While the people utilizethe light energy depending on the effectiveness of natural lighting. Fully utilizing thenatural light is of the great significance for saving the energy consumption of artificiallighting and environment protection.Sky luminance distribution is the main study objectof natural lighting.It requires a large amount of light climate data for studying the rule of skyluminance distribution accurately. If we use the traditional data analysis methods todeal with those data, it can not meet our demands inevitably.Therefore, it’s verynecessary to introduce the data mining technology into the study of sky luminancedistribution, which is very significant to promote the technology development onconstruction lighting. Light climate data in this paper comes from a light climateobserving station in Chongqing. Objective or human factors in data collection processmay cause the data quality problems, such as data loss, non-standard data format anddata duplication in original data.So it is hard to apply these data into data mining.Inorder to improve the quality of the data mining result, firstly, this paper use some datapreprocess approaches, such as data cleaning and data transformation, to handle theoriginal light climate data.It presents preprocessing approaches for the light climatedata. The task of data mining in this paper is to find the sky classification regularpattern in lager amounts of light climate data.We can identify the current sky typerapidly and accurately by the real-time sampling for light climate data.Then obtain thesky luminance distribution based on the CIE general sky luminance distribution standard and the calculation formula of luminance distribution, which provide atheoretical basis for rational utilization of natural light.Taking the light climate datafeatures and the characteristics of various classification algorithms into consideration,this paper puts forward the sky classification model for light climate data based on theneural network. In the whole process, this paper emphasizes the combination of expertopinion and practical application for handling data more accurate and more targeted.The experimental results show that the data preprocessing approaches proposed inthis paper can improve the light climate data quality effectively, and then improve thecredibility of the next data mining results. The sky classification model based on theneural network reflects a high accuracy in test samples, which gives a theoreticalsupport of studying sky luminance distribution.