A Management Analysis And Strategy Research Of Mobile Internet Base On Data Mining

With the development of3G network, the era of mobile Internet has arrived. If theenterprises of telecommunications industry want to gain a foothold in the mobileInternet war, they must not only improve the ability of its own network to link to theMobile-Network, but also need to seek a way to avoid to be pipelined by the giantInternet companies. It includes to improve the quality and to increase the quantity oftheir mobile terminal products, to develop all kinds of applications for their mobile, toset up3Gexperience environment, to promote the music, reading, video applications,games, anime, Application Centre and so on and to increase their penetration in themarket. In this strategy, however, it is difficult to win in the competition with theInternet companies which have more strength and experience in a short time. But theycan utilize the information of users, such as website address, browse predilection,browse period of time and other information to classify users whom personalizedservice can be provided. Also companies can release traffic packages that arecustomized by users’ preferences, so users can choose the most suitable one forthemselves, and carriers can cooperate with the Internet companies to get rid of uglysituation. In one word, carriers must analyze and manage the traffic produced bymobile in order to gain profit.In this background, this paper use data mining technology to develop effectivemarketing of telecom value-added business, study the characteristics of active users,their nature and habits, to provide scientific basis for business strategy decisions. Thisarticle researches on following aspects and achieved some results:1. Research on data extraction and cleaning method. There is effectiveinformation in the data which is produced by mobile phone customers when they surfthe internet by their phones. But also there is a large number of invalid data. Invalid data is undoubtedly the interference sources of data mining, invalid data cleaning andeffective information extracting are the basic works of data mining. According to thetarget of this article and the situation of data source, with relational databases as thedata source, this article uses a method which combining some kinds of extractivetechnologies, an extraction model of mobile Internet data set is put forward, togetherwith the method of information filtering to preliminary data cleansing, get a initialdata set which is satisfied for the data quality standards.2. Research on data preprocessing and discretization method. No matter howgood a data mining algorithm is, if there is no a complete and accurate training dataset, it is impossible to get a good model, so the data preprocessing technology is veryimportant in data mining. In addition, most of the data in practical application iscontinuity, but the continuous data often affect the results and the efficiency of thedata mining, therefore, most data mining tools require discrete attributes of a data set.For the above problems, this article analyzes the existing popular pretreatmenttechnology and the discretization technique, and applied to the mobile Internet dataset and got a better data set. In addition, according to the features and characteristicsof the theme, proposed an improved heuristic discretization method, it proved thatcompared with several other existing discretization algorithms, the algorithm hashigher classification accuracy by experiment.3. The establishment and optimizer of the model. The final task of data mining isto build a model. How to select the best algorithm from the numerous algorithms is aquestion for many experts and scholars, in this paper, several commonly classificationalgorithms are used on the mobile Internet data set, and through the comparativeanalysis we choose C4.5algorithm to establish the user preference model. Furtheroptimizing of this model and giving some proposals to the model by considering morefactors.4. Research on the decision tree pruning technology. The results of data miningshould be presented to the user or domain expert in the form of clear and concise. Ifthe generated rules are complex and difficult, we can say that the data mining task isfail. A well decision tree pruning algorithm should ensure the accuracy and easy tounderstand as much as possible, therefore, this article analyses the existing pruningtechnology, and proposed a pruning algorithm based on the MDL with a thresholdvalue. It proved that this algorithm can effectively reduce the size of the tree byexperiment.Overall, the new algorithms can effectively improve the overall performance of the mobile Internet traffic analysis model, the model can complete the classification ofusers, provide data support for decision makers to make marketing strategy. It hashigh application value and good market prospect.