Case Study: Aberdeen Asset Management achieves 8× faster machine-learning portfolio backtesting and improved asset-allocation decisions with MathWorks

Aberdeen Asset Management Implements Machine Learning–Based Portfolio Allocation Models in the Cloud

Aberdeen Asset Management, a global asset manager with a dedicated Solutions business, needed to improve portfolio allocation by applying machine learning to characterize complex factor relationships across equities, bonds, commodities, and property. Aberdeen turned to MathWorks tools—primarily MATLAB with toolboxes such as Statistics and Machine Learning Toolbox, Deep Learning Toolbox, Datafeed Toolbox, Parallel Computing Toolbox, and MATLAB Distributed Computing Server—to train and backtest models on more than 15 years of market data, a task too large for local PCs.

Using MathWorks software on Microsoft Azure VMs, Aberdeen developed classification models (neural networks, decision trees, SVMs), prototyped parallel code with Parallel Computing Toolbox, and ran large backtests via MATLAB Distributed Computing Server on an 80‑worker cluster. The implementation cut processing time from about 24 hours to 3 hours, supported portfolio allocation decisions used in live mandates, and increased confidence by confirming trading signals across multiple machine‑learning techniques.

Aberdeen Asset Management

Emilio Llorente-Cano

Senior Ivestment Strategist