/Data-Science

Primary LanguagePython

Data-Science/data_mining/Hidden_Markov_Model/

The goal of this project is to analyze the user's behavior on the web pages using a hidden markov model.

stack Overflow

The final code contains three folder:

  • Clustering_Users : an implementation of a k-means algorithm that should be used to classify a list of users by relying on their behaviours.
  • data : a csv file contains the data.
  • Hidden_Markov_Model: an implementation of hidden markov model
  • Markov_Matrix: extraction of markov matrix from the data.
  • test: an example of test.

This project is a simple GUI developed into python 2.7. The goal of this tool is to simplify the data mining collected from analytics database. This allows to test the random forest algorithm with the data given by the shop_match file, to perform an unsupervised clustring of global data and to make some statistical analysis ( e.g., display the histogram of a given collection).

####### recommandations ########

Before testing this tool, please make sure you have the following libraries:

  • Tkinter: for GUI

  • PIL: image processing

  • pandas: statistical analysis

  • sklearn: statistical analysis

  • kmodes: pip install kmodes from https://github.com/nicodv/kmodes

  • shop_match.bson : the data file.

####### Usage ########

python TOOL_MINING_VERSION_ALPHA_KRAYNI.py # This script is the main file.

####### Explanations ######## The "data" button is to transform a BSON file ( given by mongodb) into a csv file. The "machine learning" button is to test algorithms like random forest and to make the sensitivity analysis (e.g., study the impact of such parameter on the final model). The button "clustering" is to test the classification algorithm "K-prototypes." The statistics button is to display histograms (number of click / view by each city), cpc for each city .... Each button has various options (menu button).

stack Overflow

The help button can display a paragraph explaining the steps followed in each algorithm.