Factor Models: CAPM to Machine Learning

Module: Fintech AI | Difficulty: Advanced

CAPM

Fama-French 3-Factor

Factor Zoo

Over 400 published factors, many fail to replicate.

Replication Criteria

import numpy as np
from sklearn.linear_model import LinearRegression

class FactorModel:
    def __init__(self):
        self.model = LinearRegression()
    def fit(self, factors, returns):
        self.model.fit(factors, returns)
        self.betas = self.model.coef_
        self.alpha = self.model.intercept_
    def predict(self, factors):
        return self.model.predict(factors)
    def r_squared(self, factors, returns):
        return self.model.score(factors, returns)

Research Insight: The factor zoo crisis reveals the problem of p-hacking in finance. Many published factors are spurious. The key is to test factors out-of-sample and require economic rationale. Machine learning can discover factors but increases the risk of overfitting.