ahmdrz / spam-classifier

Here are our exercises of implementing classification algorithms.

Classification algorithms (SpamClassifier)

Here are our exercises of implementing classification algorithms in Python using sci-kit learn.

It written in Python-3.6.7. Dependencies are available in requirements.txt file. You may have to install tkinter. Follow this instruction:

$ # on the debian-based OS like Ubuntu
$ sudo apt-get install python3-tk

Image is from developers.google.com

You can see DOCUMENT.md for more information.

Docker

To run this program without installing python3 and other libraries/dependencies, you can run our docker image.

$ docker pull ahmdrz/spam-classifier:latest
$ docker run ahmdrz/spam-classifier:latest

Dataset

We used standard dataset named spambase. You can find it in dataset directory of our repository. This program support all of arff datasets that the class-label is in the last column.

Algorithms

1. kNN
2. Naive bayes
3. Decision tree
4. SVM
5. Random forest

TODO: With neural-networks

Results

The result contains the confusion matrix and the accuracy of each algorithm and will be available in the results directory.

Accuracy graph	Confusion matrix for kNN with k=6

The configuration of each classifier listed below

1. n_neighbors in kNN: 6
2. C in SVC: 2.0
3. n_estimators in RandomForest: 6
4. all others were in the default configuration.

We used confusion_matrix_pretty_print.py to generate this figure.

kNN	SVM	Naive-Bayes	Random-Forest	Decision-Tree

Authors

1. Nastaran Kiani (@Nastarankiani)
2. Ahmadreza Zibaei (@ahmdrz)