Kaggle Machine Learning Tutorials(Chapter 4)

前文提要

上章我们建立了模型来拟合数据，这章我们将检验模型。

Model-Validation(check accuracy,验证)

You've built a model. But how good is it?

In this lesson, you will learn to use model validation to measure the quality of your model. Measuring model quality is the key to iteratively improving your models.

What is Model Validation

You'll want to evaluate almost every model you ever build. In most (though not all) applications, the relevant measure(相关度量) of model quality is predictive accuracy. In other words, will the model's predictions be close(接近) to what actually happens.

Many people make a huge mistake when measuring predictive accuracy. They make predictions with their training data and compare those predictions to the target values in the training data. You'll see the problem with this approach(方法) and how to solve it in a moment, but let's think about how we'd do this first.

You'd first need to summarize the model quality into an understandable way. If you compare predicted and actual home values for 10,000 houses, you'll likely find mix of good and bad predictions. Looking through a list of 10,000 predicted and actual values would be pointless. We need to summarize this into a single metric(a system or standard of measurement).

There are many metrics for summarizing model quality, but we'll start with one called Mean Absolute Error (also called MAE). Let's break down this metric starting with the last word, error.

The prediction error for each house is:

error=actual−predicted

So, if a house cost $150,000 and you predicted it would cost $100,000 the error is $50,000.

With the MAE metric, we take the absolute value of each error. This converts each error to a positive number. We then take the average of those absolute errors. This is our measure of model quality. In plain English, it can be said as

On average, our predictions are off by about X.

To calculate MAE, we first need a model. That is built in a hidden cell below, which you can review by clicking the code button.

# Data Loading Code Hidden Here
import pandas as pd

# Load data
melbourne_file_path = '../input/melbourne-housing-snapshot/melb_data.csv'
melbourne_data = pd.read_csv(melbourne_file_path) 
# Filter rows with missing price values
filtered_melbourne_data = melbourne_data.dropna(axis=0)
# Choose target and features
y = filtered_melbourne_data.Price
melbourne_features = ['Rooms', 'Bathroom', 'Landsize', 'BuildingArea', 
                        'YearBuilt', 'Lattitude', 'Longtitude']
X = filtered_melbourne_data[melbourne_features]

from sklearn.tree import DecisionTreeRegressor
# Define model
melbourne_model = DecisionTreeRegressor()
# Fit model
melbourne_model.fit(X, y)

DecisionTreeRegressor(criterion='mse', max_depth=None, max_features=None,
                      max_leaf_nodes=None, min_impurity_decrease=0.0,
                      min_impurity_split=None, min_samples_leaf=1,
                      min_samples_split=2, min_weight_fraction_leaf=0.0,
                      presort=False, random_state=None, splitter='best')

Once we have a model, here is how we calculate the mean absolute error:

from sklearn.metrics import mean_absolute_error

predicted_home_prices = melbourne_model.predict(X)
mean_absolute_error(y, predicted_home_prices)

434.71594577146544

The Problem with "In-Sample" Scores

The measure we just computed can be called an "in-sample" score. We used a single "sample" of houses for both building the model and evaluating it. Here's why this is bad.

Imagine that, in the large real estate market, door color is unrelated to home price.

However, in the sample of data you used to build the model, all homes with green doors were very expensive. The model's job is to find patterns that predict home prices, so it will see this pattern, and it will always predict high prices for homes with green doors.

Since this pattern was derived(派生) from the training data, the model will appear accurate in the training data.

But if this pattern doesn't hold when the model sees new data, the model would be very inaccurate when used in practice.

Since models' practical(实际) value(价值) come from making predictions on new data, we measure performance on data that wasn't used to build the model. The most straightforward way to do this is to exclude some data from the model-building process, and then use those to test the model's accuracy on data it hasn't seen before. This data is called validation data.