Facebook Post Page’s Cluster

Dependences

• python (3.6)
• pandas (0.19.2)
• numpy (1.13.1)
• nltk (3.2.2)
• scipy (0.19.1)
• sklearn

Usage

Download Facebook Crawler (Facebook-Page-Crawler)

> git submodule init

> git submodule update

Download Page Posts for clustering

> python Crawler/Facebook_Page_Crawler.py 'SiteOmelete' '2017-07-01 00:00:00' '2017-07-30 23:59:59' --resume

Clustering

> python Cluster/run.py 'SiteOmelete'

Default Params

Change the information in the file parameters.json

{
    "fb_app_id": "",
    "fb_app_secret": "",
    "pca_variance_max": 0.7,
    "range_of_cluster": [10, 50],
    "tfidf_max_features": 200000,
    "tfidf_max_df": 0.1,
    "tfidf_min_df": 0.01,  
    "tfidf_ngram_range": [1, 2]    
}

Step by Step

• Data mining in text with bag of words and n-grams
• Reduction of Dimensionality with PCA
• Clustering with k-means
• Automated generation of descriptors with decision tree
• Cluster engagement analysis