The complexity and growth of financial markets and the increasing diversity of investors have
presented a demand for market analysis based on data, with intelligent data-driven approaches
to market analysis. Different conventional techniques of analysis are usually ineffective to
include the nonlinearity between financial performance indicators and behavioural features
of the performance of investors. This paper will address a machine-learning-based system to
understand investor market behaviour and categorize investor performance by a structured set
of 6000 investor records with demographic, financial, and behavioural data. The main
characteristics are age, annual income, size of investment, its past return percentage,
diversification of the given portfolio, type of investor and risk profile. Four controlled
machine learning classifiers used include Logistic Regression, Decision Tree, random Forest
and grade boosting which were implemented and tested to classify investor performance
based on the high, moderate, and low return. The findings of the experiment indicate that
Logistic Regression excelled the highest classification (99 percent) which was followed by
Gradient Boosting (95.08 percent), Random Forest (94.17 percent) and Decision Tree (88.67
percent). That analysis also determined the high level of discrimination of the Logistic
Regression and Gradient Boosting as they achieved close to perfect scores of nearly 1.00. The
results indicate the efficiency of the combination of financial and behavioural characteristics
towards predicting investor performance and prove the efficiency of ensemble and linear
models as algorithms to deal with complex investor data. The suggested framework can give
investors, financial analysts, and decision-makers a lot of insights because it allows
classifying performance correctly and then makes informed investment decisions..