← projects

Housing price prediction with DNN

Keras deep neural network regressor for housing price prediction, with a pure-NumPy 2-hidden-layer baseline for comparison.

shipped PythonKerasTensorFlowNumPyPandas

What

Two models trained side-by-side on tabular housing data:

  • Keras DNN — saved with scaler, .keras model file, metrics.json, and training history plot.
  • NumPy baseline — 2-hidden-layer network implemented from scratch; no framework, just matrix ops.

Why the NumPy baseline

Writing backprop by hand, even once, makes the framework abstraction meaningful rather than magical.

Reproducibility

Deterministic seeds set across NumPy, Python, and TensorFlow. Results are stable across runs.