🏡 Beating Boston Housing Kaggle Competition. 🏤

About Competition • About Methods • Result • Feedback

About Competition

We're going to use the Boston Housing dataset and to teach a model to predict the MEDV (median value) from the other features. The train.csv file contains the MEDV value, but the test.csv doesn't. Then we export our predictions for MEDV into a csv file, following the format in the provided sampleSubmission.csv file. The predictions will be scored against a held-out dataset.

About Methods

I've used a simple three-layer fully-connected neural network:

model4 = tf.keras.Sequential([
                              tf.keras.layers.Dense(100, activation = 'relu'),
                              tf.keras.layers.Dense(100, activation = 'relu'),
                              tf.keras.layers.Dense(1)
])

model4.compile(optimizer = tf.keras.optimizers.Adam(lr = 0.001),
               loss = 'mse',
               metrics = 'mae')

model4.fit(df, target, 
                      epochs = 1200,                     
                      verbose = 0,
                      callbacks=[PrintSmile()])

Result

Note Because that's a late submission, you'll not see me on leaderboard. So I publish result here as a screenshot. You can rerun this model and check the result.

Feedback

🤵 Feel free to send me feedback on Telegram. Feature requests are always welcome.

🧮 Check my other projects.