Objective

• Leverage analytical tools and technologies on company filings. 
• Draw insights from quarterly earnings call transcripts to generate signals and insights for the airlines sector
  

Platforms used

• Python, NLP algorithms
  

CRISIL's solution

• Data extraction: Set up scripts to convert statements to text format
  
• Data processing
  
  • Pre-processing of statements to remove stop words and special characters
  • Lexicon normalisation using lemmatisation
    
• Feature extraction
  
  • Extract meaningful word pairs(bi-grams) using tf-idf, which represent the major challenges/factors affecting the airlines sector (financial, competition, fuel-price, fares/surcharges, customer satisfaction).
    
• Variable creation
  
  • Prepared a set of independent variables pertaining to the above factors using
    - Word frequencies and sentence counts on presentation and Q&A sections
    - Sentiment on sentences by the speaker containing important factors
    - Tone of executives and analysts
• Output
  
  • Create derived variables YoY change (%) to observe the impact across time
    
• Data science
  
  • Applied dimension reduction techniques for variable reduction 
  • Created individual feature chart across timeline to draw insights
  • Built a regression model and back-tested the model
    

Client impact

• Strong correlation on adjusted returns observed with certain features 
• Feature chart across timeline to draw insights on impact to the airlines sector
• Performed sentiment analysis, reduced dimensionality, built regression model and back-tested to drive analysis
  

Questions

Looking for high-end research and risk services? Reach out to us at: