Research On Optimal Prediction Processing Model Based On Big Data

With the advent of the big data era, human society has entered a new digital era. In big data era, the basic work is data generation and collection, and the key technology is data mining. The challenge caused by the data explosive growth is the big data analysis in the ever-changing applications. It is a problem how to effectively use these data. It is also important to study the large-scale forecasting models of large-scale data using the known data.In this paper, binding data characteristic of the era of big data, firstly the background and significance of Big Data predictive processing mode will be described in detail. Then, based on the domestic and international research situation, as well as the basic machine learning and data mining algorithms, a big data forecasting and processing method is proposed.And a multimodal fusion algorithm is proposed which well combined the GBDT and the generalized linear model, and then validated by the data set. According to the big data processing model, the data set is processed.After data processing, such as data problem analysis, preprocessing, data cleaning, feature engineering, algorithm framework processing and model evaluation, the advantage of the multimodal fusion algorithm is verified,and so do the accuracy of the whole big data prediction processing mode,so the optimal prediction processing mode is obtained based on big data.Finally, Hadoop's distributed processing platform is designed and built based on the big data prediction processing model, which solves the shortcomings of slow data processing speed of the traditional processing platform. The multi-mode fusion algorithm is realized by combining Mahout platform and Hadoop platform. Layered design and the forecast processing system platform are implemented, and finally the entire big data prediction processing platform is completed.