finding-popular-consistent-topics

Covid19 is a global epidemic that affects many aspects of society and is discussed a lot on social network such as Facebook, Twitter. What topics become popular together? And these are topics that are discussed frequently or only appear in a certain period of time? The purpose of this project is finding popular consistent topics in each time period from the COVID19 Tweets dataset. This project will implement two approaches to finding popular consistent topics:

• The first approach is trying some finding frequent items set algorithms to find words appear together in many tweets in a time period.
• The second approach is apply some topic modeling methods to find latent topic in each tweets and then find topics become together in a time period by applying some finding frequent items set algorithms.

How to run

Requirements

python3.6

pip install -r requirements.txt

Run finding similar item sets (first approach)

python finding_similar_items.py [--pre_process_type remove_twitter_account] [--alg fpg] [--min_sup 0.01] [--min_conf 0.01]

• pre_process_type: method when applying pre-process tweet content
  • remove_url
  • remove_twitter_account
  • remove_url_replace_twitter_account
  • remove_twitter_account_replace_url
  • replace_twitter_account_and_url
• alg: finding similar item set algorithm
  • fpg: Frequent Pattern Growth algorithm
  • pcy: PCY algorithm
  • apriori: A-Priori algrithm
• min_sup, min_conf: support and confidence parameter of fiding similar item set algorithm
• The result in output folder

Run finding popular topics (second approach)

python finding_popular_topic.py [--pre_process_type remove_twitter_account] [--alg_similar fpg] [--alg_topic LDA] [--num_topic 20]

• pre_process_type: same as finding similar item sets
• alg_similar: finding similar topics algorithms
  • fpg: Frequent Pattern Growth algorithm
  • pcy: PCY algorithm
  • apriori: A-Priori algrithm
• alg_topic: algorithm for finding latent topics from each tweets
  • LDA: Latent Dirichlet Allocation
  • BTM: Biterm topic model
• num_topic: number of topic
• The results in output folder