Machine Learning and Data Science

In lecture, we talked about how Machine Learning just boils down to making predictions about the world, given adequete data. We also learned about a variety of different techniques for the machine learning process, namely data pre-processing and splitting data into a training/test split.

Slides from lecture are available here.

About Pandas and Sci-Kit Learn

Recall from lecture that Pandas is a library that gives us a framework for loading and manipulating data. Data is stored in a spreadsheet-like format of a DataFrame, which we can initialize from a .csv file (along with a list of dictionaries or by inserting rows into programatically). In addition to Pandas, we will also make use of .

Analyzing Cereal

In lecture, we took a look at the Cereal Dataset from Kaggle.

Working With Files

We used the following script to read in the "cereal.csv" file with Pandas:

import pandas as pd
import matplotlib.pyplot as plt


def get_cereal_df(input_path):
    return pd.read_csv(input_path)


def visualize_data(df):
    plt.scatter(df['calories'], df['sugars'])
    plt.show()


if __name__ == '__main__':
    df = get_cereal_df('cereal.csv')
    visualize_data(df)

Note that the code under the if statement is only run when we execute the script via the command line (i.e. python3 data.py). This is to ensure that we can later import functionality from this module without running all the code in the script.

Training a Machine Learning Model

Finally, we trained a machine learning model named KNeighborsRegressor to predict the rating of a cereal, given it's sugar and calories as features. The model is a variant of the K-Nearest Neighbours classifier discussed in class. However, the model performs a regression task of predicting a continuous value, rather than a discrete one.

We first import the data points via get_cereal_df, partition the data into a training and testing splits, and then train the classifier:

from sklearn.neighbors import KNeighborsRegressor
from sklearn.model_selection import train_test_split
from data import get_cereal_df


if __name__ == '__main__':
    df = get_cereal_df('cereal.csv')

    train_features, test_features = train_test_split(df[['sugars', 'calories']], test_size=0.2)
    train_labels, test_labels = train_test_split(df[['calories']], test_size=0.2)

    model = KNeighborsRegressor(n_neighbors=9)
    model.fit(train_features, train_labels) # ---> does the training!

    print(model.score(test_features, test_labels))

Our final line prints the model's evaluation score, which you can read about here.

Conclusion

There are many different third-party tools and frameworks that make machine learning easy in Python. We discussed in class how they use Cython to speed up the performance of the code, and this will become more apparent in the following week's lectures, when we look into Natural Language Processing and Deep Learning!