Extract Data from top 20 publishers & the recent 3 years (i.e., 2020-2022)

Please refer to file: 2020to2022data.csv in dataset

Split the Dataset into small parts and translate using google API. Combine the translated files into one Create Column "tran_headline" and "tran_content"

Please refer to file: 2020to2022transdata.csv in dataset

Remove null / poorly translated data and perform text preprocessing

Adding ESG categories and sentiment labels to the dataset With reference to model:

Sentiment: https://huggingface.co/cardiffnlp/twitter-roberta-base-sentiment

Esg Categories Classification: https://huggingface.co/yiyanghkust/finbert-esg-9-categories

Please refer to Release: 20_22ESGSentidata.csv in dataset

Please check the folder "5_Code and Data for Keyword Extraction" Code can be found in "5_Code and Data for Keyword Extraction/Code" Dataset extracted by Keywords and Graphs can be found in "5_Code and Data for Keyword Extraction\Keyword datasets"

Align the dataset with HSI index to generate further insight Please find the aligned dataset in folder "6_Groupby Dataset"

Build regression and classification models with ESG sentiment and label counts. We do it in 2 Approach: For predicting close price using close price of HSI and senti label, check "7_Build Predictive Model\LSTM_Predict close price"

For predicting Daily Return and Classification of positive/negative return, check "7_Build Predictive Model\Predict Daily Return"