Our algorithms are built on decades of academia research.
Machine Learning for Recession Prediction and Dynamic Asset Allocation by David Rapach, Jack Strauss, Guofu Zhou, Jun Tu. This paper proposes a novel machine learning approach to predict recessions and allocate assets dynamically. The authors use support vector machines (SVMs) to forecast the probability of a recession based on a large set of macroeconomic and financial predictors. They then use the recession probability as an input to a dynamic risk budgeting framework that allocates assets between risky and safe portfolios. The authors show that their approach outperforms several benchmarks in terms of Sharpe ratio and maximum drawdown.
A Machine Learning Approach to Tactical Asset Allocation by Rui Pedro Brito. This paper applies supervised machine learning algorithms to the tactical asset allocation problem. The author uses random forests and gradient boosting machines to select portfolio assets based on macroeconomic indicators such as GDP growth, inflation, interest rates, etc. The author compares the performance of his machine learning models with traditional methods such as mean-variance optimization and equal-weighting. He finds that his models generate higher returns and lower volatility than the traditional methods.
Dynamic Asset Allocation with Deep Neural Networks by Jiaqi Chen, Xiang Li, Zhenyu Wu. This paper presents a deep neural network based approach for dynamic portfolio optimization that can handle nonlinearities and interactions among assets. The authors use a feed-forward neural network to model the expected return and covariance matrix of a large number of assets. They then use a recurrent neural network to capture the time-varying dynamics of asset returns and volatilities. They solve the portfolio optimization problem using stochastic gradient descent with backpropagation. The authors show that their approach outperforms several benchmarks in terms of risk-adjusted returns and turnover.
Asset Allocation via Machine Learning by Jiaqi Chen, Xiang Li, Zhenyu Wu. This paper documents a novel machine learning-based numerical framework to solve static and dynamic portfolio optimization problems with a large number of assets. The authors use convolutional neural networks to extract features from high-dimensional asset data such as prices, returns, volatilities, etc. They then use feed-forward neural networks to approximate the optimal portfolio weights for different risk preferences and constraints. They also propose an online learning algorithm that updates the portfolio weights in real time based on new data. The authors demonstrate that their framework can handle various portfolio optimization problems such as mean-variance optimization, minimum variance optimization, maximum diversification optimization, etc.