series

Financial machine learning from zero

A beginner-friendly series on machine learning in finance, built around the problems banks actually run models on: credit scoring, fraud detection, and anti-money-laundering analytics, with market prediction as a supporting track. Every part with code comes with a runnable companion folder. No GPU needed.

for
People who know some Python and have never trained a model on financial data.
published
2 parts · 49 min of reading
planned
5 more
code
github.com/prashant-fintech/financial-ml-from-zero ↗
Start with part 1

parts

  1. 01

    The vocabulary, the traps, and the metrics that matter

    The first post in a beginner-friendly series on machine learning in finance, built around the problems banks actually run models on: credit scoring, fraud detection, and anti-money-laundering analytics, with market prediction as a supporting track. No code yet. Instead, the things I wish every newcomer knew before training a model on financial data: why it breaks the textbook assumptions, the vocabulary in every model notebook, the mistakes that make models look brilliant in development and fail in production, and a reference table of the metrics that matter, with which direction is good for each.

    22 min readcode ↗

  2. 02

    Tensors, time-ordered batching, and why leakage hides in .shift()

    The second post in the financial machine learning series, and the first with code. It turns a Freddie Mac-style mortgage panel into PyTorch tensors one named step at a time, then breaks the pipeline on purpose with eight one-line mistakes and measures each one. Some push validation PR-AUC from 0.31 to 0.96. Others barely move it, which is worse, because nothing looks wrong. It ends with the checks that catch every one of them before a reviewer does.

    27 min readcode ↗

  3. 03

    A first probability-of-default baselineplanned

    Logistic regression in PyTorch on the same tensors, a naive benchmark to beat, a loss that handles class imbalance, and an honest report including calibration. The test split gets used, once.

  4. 04

    Purged walk-forward validationplanned

    The splitting scheme every later post uses, with the purge and embargo explained properly.

  5. ··

    Fraud detection with late-arriving labelsplanned

    Training and evaluating when a fraud label only exists once a customer notices and disputes.

  6. ··

    The anomaly layer of an AML pipelineplanned

    How to measure it when almost nothing is labelled.

  7. ··

    A short market trackplanned

    A return-prediction baseline, and the ranking and strategy metrics applied honestly.

Follow new parts via RSS.