Investment time series prediction using a hybrid model based on RBMs and pattern clustering

Abstract

The concept of internet finance has attracted increasing attention in recent years. As a result, more and more online peer-to-peer (P2P) lending platforms have been established at home and abroad. It is actually meaningful to predict investment amounts of online lenders in the following period. In this paper, we propose a Hybrid Investment Prediction Model (HIPM), an effective non-linear prediction method, which involves spectral clustering with a novel distance measure to discover similar characteristics of investment trends and Restricted Boltzmann Machine (RBM) models to forecast the future points with a particular initialization according to the investment pattern of each lender. The prediction accuracy of HIPM on a data set containing thousands of lenders collected from PPDAI website, a P2P lending platform in China, outperforms traditional prediction methods including Autoregressive Integrated Moving Average (ARIMA) and Support Vector Machine (SVM) models.