Classification Algorithims on CSV Datasets

Description of the Goal and how to Execute

The goal of the project is to develop and evaluation classifiers for a dataset.

Given a new person with some of the attributes from the table, I will be predicting whether that person will make more or less than 50k per year.

This dataset has a little more than 7,000 instances and has seven attributes.

Run this program by entering the following: python classify.py > ../output/output_file_name

• This will generate an output file with all of the classifier information in a file name of your choosing.
• The visualization part of this project can be found in the graphs folder once the program has run to completion.
• The "output_file_name" part of the program execution is not necessary. It is there because there is a matplotlib warning in the output that I can not find a solution for. Until I find a solution, the "prettiest" way to see the output is if you direct the output to a file.

Requirements - Download and Install:

1. numpy

   pip install numpy
   

2. matplotlib

   pip install matplotlib
   

3. tabulate

   pip install tabulate
   

Details about the Sample Dataset

For my project I will be using a dataset with 20,000 instances that will help predict whether a particular person makes more than 50k a year based on particular qualities. The descriptions of these attributes are defined below.

1. Age

   The current age of each person when the data was collected.

2. Degree

   The highest level of education that each person had when the data was collected. This can be High School, Bachelors, Masters, Doctorate, College Drop Out, Associate, Middle School, Elementary, or Professional School.

3. Marital Status

   The current marital status of each person when the data was collected. This can be Never Married, Married, Divorced, Spouse Absent, Widowed, or Seperated.

4. Ethnicity

   The ethnicity of each person that had their data collected. This can be White, Black, Native American, or Asian.

5. Gender

   The gender of each person that had their data collected. This can be Male or Female.

6. Country

   The current country that each person resided at the time of the data was collected. This could be the United States, the Philippines, Puerto Rico, Mexico, the Dominican Republic, Portugal, Canada, Taiwan, Cuba, or Jamaica.

7. Salary: The attribute that we will classify on

   The current salary that each person made at the time the data is collected. This will be the attribute that we are classifying on. This could be either >50k or <=50k.

Note: Every attribute (besides age) are categorical. This meant that I needed to go through the entire data set and discretize it, making a temporary dataset that has continuous attributes that represent categorical data.

File Names and Descriptions

constants.py

Only for constant variables. No functions please.

file_system.py

Only for reading/writing files.

math_utils.py

Only for math. Just math math math math.

table_utils.py

For manipulating a 2d array (table).

homework_util.py

Put functions in here that are specific to a homework assignment.

naive_bayes.py

This is the naive_bayes module.

knn.py

Contains functions pertaining to KNN

classifier_util.py

This is utilities that can apply to any classifier. For example: accuracy, error rate, ect.

partition.py

If you're splitting up a table, put it here

decision_tree.py

Functions for decision tree stuff

random_forest.py

Functions for random forest stuff

Output

===========================================
Rows with NAs have been removed.
===========================================

===========================================
Data visualization complete.
===========================================

===========================================
K-Nearest Neighbors
===========================================
	Random Subsample
		Accuracy = 0.784615384615, error rate = 0.215384615385
	Stratified Cross Folds (5)
		Accuracy = 0.8, error rate = 0.2

===========================================
Confusion Matrix
===========================================
========  ===  ===  =======  =================
  Salary    1    0    Total    Recognition (%)
========  ===  ===  =======  =================
       1    0    8        8                  0
       0    0   32       32                100
========  ===  ===  =======  =================

===========================================
Naive Bayes
===========================================
	Random Subsample
		Accuracy = 0.757692307692, error rate = 0.242307692308
	Stratified CrossFolding
		Accuracy = 0.8, error rate = 0.2

===========================================
Confusion Matrix
===========================================
========  ===  ===  =======  =================
  Salary    1    0    Total    Recognition (%)
========  ===  ===  =======  =================
       1    2    6        8              25
       0    2   30       32              93.75
========  ===  ===  =======  =================

===========================================
Decision Tree
===========================================
IF degree == 1 THEN salary = 0
IF degree == 2 THEN salary = 0
IF degree == 3 THEN salary = 1
IF degree == 4 THEN salary = 1
IF degree == 5 THEN salary = 0
IF degree == 6 THEN salary = 0
IF degree == 7 THEN salary = 0


Stratified CrossFolding
	Accuracy = 0.825, error rate = 0.175



===========================================
Confusion Matrix
===========================================
========  ===  ===  =======  =================
  Salary    1    0    Total    Recognition (%)
========  ===  ===  =======  =================
       1    2    6        8             25
       0    1   31       32             96.875
========  ===  ===  =======  =================

===========================================
Random Forest
===========================================
N = 3000 M = 215 F = 2
	Accuracy = 0.8

===========================================
Confusion Matrix
===========================================
========  =======  ======  =======  =================
Salary      <=50K    >50K    Total    Recognition (%)
========  =======  ======  =======  =================
<=50K           8       0        8                100
>50K            2       0        2                  0
======== ======= ====== ======= =================

Sources

The dataset is from: http://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.data