Facebook Event Analysis and Interested Count Predictor

This is a project for analysis of Facebook Events and for getting interested count prediction.A Random Forest model was used for training on a large dataset of ~1000 events.Feature engineering,Data cleaning, Data selection and many other techniques were used for this task.

Tools Used

• Python 2.7
• Pandas
• Sklearn
• NumPy
• Seaborn
• Pickle
• NLTK
• Enchant

How to run on local machine:

1. Clone this repo

   $ git clone https://github.com/techedlaksh/fbevent-prediction
   $ cd fbevent-prediction

2. Create new virtual environment

   $ sudo pip install virtualenv
   $ virtualenv venv
   $ source venv/bin/activate
   $ pip install -r requirements.txt
    ```

3. Website

   $ python app.py &

Code Details

Below is a brief description for the Code files/folder in repo.

d3-charts/

This folder contains ipython notebook which were used to clean the data for individual d3-chart & pre-processed data for those charts.

DataPre-Processing.ipynb

The script fetches data from MongoDB Client in which data was already saved by another script. The pre-processing of data is done for plotting data on geo-map, generating a word frequency chart and popular categories by different metrics which are showcased using donut chart.

$ jupyter DataPre-Processing.ipynb

data/

This folder contains ipython notebook which contain implementation for graph api of facebook to extract events data from facebook.

DataGraphApi.ipynb

The notebook contains implementation of graph api which is used to to extract ids of ~1000 and store it in ids.json then extract all the data of each event and store it as collection in MongoDB.

Note : The raw data fetched from graph api resides in data/precog_db_data. data/precog_fb_data/events.bson has the data and data/precog_fb_data/events.metadata.json contains the metadata information about the data.

model/

This folders contains scripts used for training,tuning model and getting the prediction results.

Prediction-model.ipynb

This notebook pre-processes all the data to appropriate form which can be used as a training for our prediction model. Random Forest model is also inside this notebook and is used to get the prediction from the preprocessed data.

$ jupyter Prediction-model.ipynb

Note : The trained model with ~98% accuracy has been exported in this folder as classifier.pkl which can be imported anywhere to be used.

static/

This folders contain all the static file such css, data for charts and donut chart script to be served in flask

templates/

This folder contains all those html files which is used to be served as frontend of the flask website which is hosted on fbevent-laksh.herokuapp.com