sindhri / boston_housing_prediction

Use 13 features to predict Boston Housing Price, MLP with cross validation

Boston Housing Price prediction

Use 13 features to predict Boston Housing Price, MLP with cross validation
Credit: Deep Learning with Python by Jason Brownlee
Data source: https://raw.githubusercontent.com/jbrownlee/Datasets/master/housing.data

data size: 506 x 14

Feature list:
1.CRIM: per capita crime rate by town.
2.ZN: proportion of residential land zoned for lots over 25,000 sq.ft.
3.INDUS: proportion of non-retail business acres per town.
4.CHAS: Charles River dummy variable (= 1 if tract bounds river; 0 otherwise).
5.NOX: nitric oxides concentration (parts per 10 million).
6.RM: average number of rooms per dwelling.
7.AGE: proportion of owner-occupied units built prior to 1940.
8.DIS: weighted distances to five Boston employment centers.
9.RAD: index of accessibility to radial highways.
10.TAX: full-value property-tax rate per$10,000.
11.PTRATIO: pupil-teacher ratio by town.
12.B: 1000(Bk−0.63)2where Bk is the proportion of blacks by town.
13.LSTAT: % lower status of the population.

The last column is the variable to be predicted 14.MEDV: Median value of owner-occupied homes in$1000s.

Baseline model: simple MLP with 2 layers

MSE = -46.58 (32.97)

With standarization MLP baseline 2 layers

MSE = -25.88 (23.12), better

Deeper net (3 layers):

MSE = -26.01 (31.76), not better, run again -20.72 (22.95), -25.61 (29.38)

Wider net with 20 first layer neurons(2 layers)

MSE = -21.73 (23.52), better

conclusion, a wider 2-layer net has the best performance with the final MSE -21.73