Curriculum Learning in Deep Neural Networks for Financial Forecasting

Abstract

For any financial organization, computing accurate quarterly forecasts for various products is one of the most critical operations. As the granularity at which forecasts are needed increases, traditional statistical time series models may not scale well. We apply deep neural networks in the forecasting domain by experimenting with techniques from Natural Language Processing (Encoder-Decoder LSTMs) and Computer Vision (Dilated CNNs), as well as incorporating transfer learning. A novel contribution of this paper is the application of curriculum learning to neural network models built for time series forecasting. We illustrate the performance of our models using Microsoft’s revenue data corresponding to Enterprise, and Small, Medium & Corporate products, spanning approximately 60 regions across the globe for 8 different business segments, and totaling in the order of tens of billions of USD. We compare our models’ performance to the ensemble model (of traditional statistics and machine learning) currently being used by Microsoft Finance. Using this in-production model as a baseline, our experiments yield an approximately 30% improvement overall in accuracy on test data. We find that our curriculum learning LSTM-based model performs best, which shows that one can implement our proposed methods without overfitting on medium-sized data.