Stock News Sentiment Analysis

Stock News Sentiment Analysis/Dataset/README.md

@@ -0,0 +1,11 @@
+# Dataset:
+https://www.kaggle.com/datasets/avisheksood/stock-news-sentiment-analysismassive-dataset
+
+# Columns:
+Sentiment: 0 represents a negative/neutral sentiment and 1 represents a positive sentiment.
+
+Sentence: The text upon which sentiment analysis is to be performed.
+
+# About this file:
+0 represents that the news is negative or neutral (Therefore the stock will likely go down)
+1 represents that the news is positive (Therefore the likely stock will go up)

Stock News Sentiment Analysis/Model/README.md

@@ -0,0 +1,79 @@
+# Stock News Sentiment Analysis
+
+**PROJECT TITLE**
+
+**GOAL**
+
+The goal is to perform sentiment analysis on stock market news. It is a binary classification problem, where 
+- label==1 signifies positive sentiment, and, 
+- label==0 signifies neutral or negative sentiment.
+
+**DATASET**
+
+https://www.kaggle.com/datasets/avisheksood/stock-news-sentiment-analysismassive-dataset
+
+**DESCRIPTION**
+
+
+
+**WHAT HAVE I DONE**
+
+- Removed stopwords, punctuations, made text lowercase, and lemmatized text to base form.
+- Made WordClouds for:
+    - The entire dataset
+    - Words featured in positive sentiments
+    - Words featured in negative sentiments
+- Vectorized text using TF-IDF vectorizer
+- Applied classification on a dataset with:
+    - Parameter: TF-IDF vectors
+    - Label: Encoded sentiment
+
+**MODELS USED**
+
+Classification algorithms like:
+- Logistic regression
+- Naive-Bayes classifier
+- SVM classifier
+- Random Forest Regressor
+
+**LIBRARIES NEEDED**
+
+- numpy
+- pandas
+- matplotlib
+- nltk
+- textblob
+- wordcloud
+- sklearn
+- tensorflow
+
+**VISUALIZATION**
+- Sentiment distribution (1 and 0)
+![Alt text](../Images/Sentiment_distribution.png)
+- WordCloud for overall dataset
+![Alt text](../Images/WordCloud.png)
+- WordCloud for negative sentiment text only
+![Alt text](../Images/WordCloud_negative.png)
+- WordCloud for positive sentiment text only
+![Alt text](../Images/WordCloud_positive.png)
+
+**ACCURACIES**
+
+Ranking models based on accuracy:
+
+- DistilBert: 54.70%
+- Naive-Bayes: 54.42%
+- SVM: 53.67%
+- Random Forest Classifier: 53.86%
+- Logistic Regression: 53.17%
+
+
+**CONCLUSION**
+
+All used models give a similar performance while using TF-IDF vectorization. 
+
+**YOUR NAME**
+
+- Name:       Shobhit Bandhu
+- College:    JU B.Prod '27
+- LinkedIn:   https://www.linkedin.com/in/shobhit-bandhu/