Case study 02

Applied analytics · Transportation

Houston METRO Bus Delay Prediction

A classification-based data science project exploring which service and route conditions are most associated with late bus arrivals in Houston.

  • Python
  • pandas
  • scikit-learn
  • EDA
  • Classification
2
arrival classes modeled
5
analysis stages
HTX
local decision context

01 · Business question

What needed to change?

Late arrivals affect rider trust and the usability of public transportation. The project asked whether available Houston METRO service data could be cleaned and structured well enough to classify an arrival as late or on time.

02 · Approach

From ambiguity to a system.

  1. 01

    Defined a binary target that separated late arrivals from on-time service.

  2. 02

    Cleaned the source data and reviewed missing, inconsistent, and extreme values.

  3. 03

    Used exploratory analysis to identify temporal and operational patterns associated with delay.

  4. 04

    Selected candidate features and compared classification behavior against a simple baseline.

  5. 05

    Evaluated the model with emphasis on interpretability, error tradeoffs, and practical limitations.

03 · Analytical flow

A repeatable path from question to action.

  1. 01METRO data
  2. 02Cleaning
  3. 03Exploration
  4. 04Feature selection
  5. 05Model evaluation

04 · Outcome

A local public-data problem translated into an interpretable prediction workflow.

  • Created a repeatable notebook workflow from raw public data to an evaluated classification result.
  • Connected model output to a real operational question rather than treating prediction as an isolated exercise.
  • Documented limitations and the additional service data needed for a stronger production model.

Reflection

The project reinforced that a useful model begins with a defensible target definition and careful data preparation—not the most complex algorithm available.
Next case study · 03Trusted KPI System