Research On Network Data Processing Based On Big Data Analysis Platform

With the continuous development of information technology and the popularization of com-puter applications,more and more people enjoyed the great achievements brought by the devel-opment of science and technology.Big data technology changed the life of people for the basic necessities.For example social networking of Facebook recommended friends to us and other applications like the online shopping of Alibaba’s Group were in our lives.With the arrival of the wave of big data,machine learning technologies based on big data aroused wild attention in academia and industry,such as the famous project Alphago go by Google,defeated many of the world’s famous go master.However,at present,there are still many problems in the develop-ment of big data technology,and we aimed at how to obtain the data information of interest in a certain situation in a large amount of network data,this paper studies the network data analysis and processing technology based on large data in two different network scenarios and the main works and innovations were as follows:1.These was network-trade data related to item recommending system trained through the popular big data technology such as Hadoop platform meanwhile giving strategies to take advantage of the effective information in the data.In the case of item trade prediction,the predicting system with multi-mode information got trained and with an algorithm of Alternating Direction Method of Multipliers,it had a better performance than the algorithm of Alternating Least Squares.2.With the data of video game,an application of depth enhancement learning in game AI is visually displayed.And then through analysis of the model input,the complexity of the original algorithm,a "critical area" training strategy was put forward and achieved a better per-formance,which leads to the reduction of the original image redundancy training and increase of the proportion of effective information in game image.An introduction of the evolutionary neural network algorithm in the game AI model was conducted through the comparison the deep reinforcement learning game AI.Meanwhile by analyzing the obstacles in the training process,a fast training strategy was put forward and could reduce training time,which led to a flexible change of the initial training conditions by judging the number of candidates.