As we know from our previous lessons, we should split our data into two types; training and testing. So far, we’ve just been using all of our data when making our models. Thankfully, sklearn can help us out a lot here. Make sure to grab the dataset from classroom and let’s dive in.

As always, load in your data and our libraries,

import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline

df = pd.read_csv("usedcarprices.csv")

Let’s visualize our data so we can tell what type of model we should use;

plt.scatter(df.Age, df.Sell_Price)

plt.scatter(df.Mileage, df.Sell_Price)

You should see two graphs with a negative correlation for both. That make’s logical sense; we should probably use Linear Regression. This time though, we’re going to perform our test/training split.

Let’s separate our input and output data;

X = df[['Mileage', 'Age']]
y = df['Sell_Price']

Then, all we need to do is import and call the method to do the split for us,

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.3)

What we just did was tell Python we want to separate our training data into 70% training data, 30% test data. We can call each new variable to see what’s inside/len() to see the length of each. From here, we can fit our model with the training data, and predict using the testing data;

from sklearn.linear_model import LinearRegression
reg = LinearRegression()
reg.fit(X_train, y_train)

test_vals = reg.predict(X_test)

Make sure to call your y_test values to compare them against your predicted values. One more parameter you should be aware of is the random_state parameter, which you can pass into your training split like in the example below to always make sure your training/testing split is the same every time it’s ran;

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.3, random_state=10)