SpamOrHam

Spam Email Classification with Decision Trees in Python

Table of Contents

• Contributors
• Abstract
• Brief Overview of Decision Trees
• Explanation of Data Wrangling Process
• Explanation of Data Analysis Process

Contributors

• Shon Inouye
• Ashriful Dulla
• Bryan Daetz
• Kevin Wang
• Jun Ki Kwon (Peter)

Abstract

This project focuses on the classification of emails as either spam or non-spam using decision trees. The data set for this project contains a list of spam and non-spam emails with their attributes in terms of frequency of certain words and characters. We used 75% of the data set to train the decision tree and applied that model on the remaining 25% of the data set to test the model accuracy. When training our decision tree, we focused on the importance of having minimal false positives as possible (good email marked as spam).

Overview of Decision Trees

For our model we used Decision Trees. A decision tree is a decision support tool that uses a tree-like graph or model of decisions and their possible consequences, including chance event outcomes, resource costs, and utility. It is one way to display an algorithm. We divided our data set into a training set and test and used decision tree classification algorithm in order to determine if an email classifed as spam or non-spam.

• Training Set - This component is used to train the algorithm to discover predicitve relationships between the dependent variables and independent variable. In our example we used 75% of the data set (chosen randomly).
• Test Set - This component is what is left out of the algorithm and used to asses how well our decision tree performs on data it hasn't seen yet. In our case the rest of the data set, so the 25% that was not chosen to be in the training set

We applied our algorithm to the test set to get a 92% accurancy of our model.

Data Wrangling

In order match the values in the data set to their corresponding labels, we needed to create a list of column names. The names are given on the data set description for this project. We took these labels and set it as a list of columns in our script.

columns = ['word_freq_make','word_freq_address','word_freq_all','word_freq_3d','word_freq_our','word_freq_over','word_freq_remove',
            'word_freq_internet','word_freq_order','word_freq_mail','word_freq_receive','word_freq_will','word_freq_people',
            'word_freq_report','word_freq_addresses','word_freq_free','word_freq_business','word_freq_email','word_freq_you',
            'word_freq_credit','word_freq_your','word_freq_font','word_freq_000','word_freq_money','word_freq_hp','word_freq_hpl',
            'word_freq_george','word_freq_650','word_freq_lab','word_freq_labs','word_freq_telnet','word_freq_857','word_freq_data',
            'word_freq_415','word_freq_85','word_freq_technology','word_freq_1999','word_freq_parts','word_freq_pm',
            'word_freq_direct','word_freq_cs','word_freq_meeting','word_freq_original','word_freq_project','word_freq_re',
            'word_freq_edu','word_freq_table','word_freq_conference','char_freq_;','char_freq_(','char_freq_[','char_freq_!',
            'char_freq_$','char_freq_#','capital_run_length_average','capital_run_length_longest','capital_run_length_total']

Data Analysis

This is the decision tree.

We found the 10 most important predictors for our Decision Tree Classifier algorithm. It seems that the most important predictor by far is the use of exclamation marks.

Predictor	Importance
char_freq_!	0.334817369007
word_freq_over	0.159363276944
char_freq_[	0.0890565948252
word_freq_money	0.055601149615
capital_run_length_longest	0.0418640694781
word_freq_addresses	0.0336232868801
capital_run_length_average	0.0288015387955
char_freq_#	0.0261924346351
word_freq_hpl	0.0260389190479
word_freq_3d	0.0238941385947

Here are the accuracy and error rate of our algorithm on the test set we created:

• Accuracy: 0.920938314509
• Error Rate: 0.0790616854909

Here is the breakdown of our correct and incorrect classifications from the test set:

	Was Not Spam	Was Spam
Predicted Not Spam	633	48
Predicted Spam	43	427