Abstract:Quantitative investment is a fundamental financial task that highly relies on accurate stock prediction and profitable investment decision making. Despite recent advances in deep learning (DL) have shown stellar performance on capturing trading opportunities in the stochastic stock market, we observe that the performance of existing DL methods is sensitive to random seeds and network initialization. To design more profitable DL methods, we analyze this phenomenon and find two major limitations of existing works. First, there is a noticeable gap between accurate financial predictions and profitable investment strategies. Second, investment decisions are made based on only one individual predictor without consideration of model uncertainty, which is inconsistent with the workflow in real-world trading firms. To tackle these two limitations, we first reformulate quantitative investment as a multi-task learning problem. Later on, we propose AlphaMix, a novel two-stage mixture-of-experts (MoE) framework for quantitative investment to mimic the efficient bottom-up trading strategy design workflow of successful trading firms. In Stage one, multiple independent trading experts are jointly optimized with an individual uncertainty-aware loss function. In Stage two, we train neural routers (corresponding to the role of a portfolio manager) to dynamically deploy these experts on an as-needed basis. AlphaMix is also a universal framework that is applicable to various backbone network architectures with consistent performance gains. Through extensive experiments on long-term real-world data spanning over five years on two of the most influential financial markets (US and China), we demonstrate that AlphaMix significantly outperforms many state-of-the-art baselines in terms of four financial criteria.

Multiple Financial Analyst Opinions Aggregation Based on Uncertainty-Aware Quality Evaluation

A Multiperiod Multiobjective Portfolio Selection Model With Fuzzy Random Returns for Large Scale Securities Data

A Biobjective Optimization Model for Expert Opinions Aggregation and Its Application in Group Decision Making.

Study on processing of risk factors in risk assessment of electric power investment

Fairness-aware large-scale collective opinion generation paradigm: A case study of evaluating blockchain adoption barriers in medical supply chain

Unsupervised Opinion Aggregation -- A Statistical Perspective

Model Aggregation for Risk Evaluation and Robust Optimization

Expert Aggregation for Financial Forecasting

A New MOPSO to Solve a Multi-Objective Portfolio Selection Model with Fuzzy Value-at-Risk.

A Novel Hybrid ICA-FA Algorithm for Multiperiod Uncertain Portfolio Optimization Model Based on Multiple Criteria

Enhancing Investment Analysis: Optimizing AI-Agent Collaboration in Financial Research

A novel financial risk assessment model for companies based on heterogeneous information and aggregated historical data

Quantitative Stock Investment by Routing Uncertainty-Aware Trading Experts: A Multi-Task Learning Approach

Building a Fuzzy Multi-Objective Portfolio Selection Model with Distinct Risk Measurements

Quality Enhancement by Weighted Rank Aggregation of Crowd Opinion

Adaptive large-scale group interactive portfolio optimization approach based on social network with multi-clustering analysis and minimum adjustment

Uncertainty-Aware Portfolio Management With Risk-Sensitive Multiagent Network

Risk measurement of aggregation approaches in multiple attribute decision making under uncertain information

Application and Comparison of Multiple Machine Learning Models in Finance

The Value of Performance Weights and Discussion in Aggregated Expert Judgments