Housing price prediction with DNN
Keras deep neural network regressor for housing price prediction, with a pure-NumPy 2-hidden-layer baseline for comparison.
What
Two models trained side-by-side on tabular housing data:
- Keras DNN — saved with scaler,
.kerasmodel 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.