UNBALANCED BIG DATA CLASSIFICATION BASED ON IMPROVED RANDOM FOREST ALGORITHM

Big data analytics has developed rapidly in recent years and data mining has been a positive driver for development in all areas, but data in many areas is grossly unbalanced, and there are still many limitations to current research on classifying big data. To solve this problem, the study uses the K -means algorithm based on class distinction to approximately reduce the dimensionality of the data, and the untracked Kalman filter (UKF) algorithm with an adaptive traceless Kalman filter (Sage-Husa) to reduce the noise of the data. The noise -reduced and dimension -reduced data were obtained to improve the random forest algorithm (K-U-S-H-RF). However, during the study of classifying low -dimensional unbalanced data using K-S-H-RF, it was found that the random forest algorithm did not take account of the actual step-by-step of the data set and was not effective in classifying the data. For this reason, the study introduced cost sensitivity, cost error calculation for decision trees as well as voting. Random forest is parallelized with MapReduce idea to achieve optimum of K-S-H-RF. Then the study constructs an imbalanced big data classification model based on improved random forests. The model can effectively classify unbalanced big data and provide a new path for big data application in more fields, which has a positive effect on the development of the big data era.