logfile.txt

-Import train data
https://raw.githubusercontent.com/dD2405/Twitter_Sentiment_Analysis/master/train.csv
-View the train data in the form of a table, check for the values and the headings
-Import test data
https://raw.githubusercontent.com/dD2405/Twitter_Sentiment_Analysis/master/test.csv
-View test data in the form of a table, check for the values and the headings



PREPROCESSING:

-Combine the train and the test data rows for the preprocessing
-Clean the twitter handles from the text and put it in a separate column
-Clean every punctuations except for text and hashes
-Remove Short words (length <= 3)
-Create tokens i.e. create another pandas object each having a list of above words from each line
-Stemming : Removing ing,ly,ed from the words
-Join the token words into string back in the 'Tidy Tweets' column




DATA VISUALISATION:
 
>> Generate Wordcloud do for bothe positive and negative labels
     -Join all the rows which are labeled positive into a variable 
      from tidy tweets column
      
     -Create a wordcloud
     Black Background:
     http://clipart-library.com/image_gallery2/Twitter-PNG-Image.png
     White Background:
     https://static01.nyt.com/images/2014/08/10/magazine/10wmt/10wmt-jumbo-v4.jpg?quality=90&auto=webp
     
     
>> Create list of positive and negative hashes(words) in the tweets
   Plot top twenty frequency words in both in the form of bargraph
   
   
   
   
FEATURE EXTRACTION:
- Create bag of words dataframe
- Create TF-IDF dataframe




SPLITTING DATA INTO TRAIN AND TEST




APPLYING MACHINE LEARNING MODELS
-Logistic Regression : predict probability, f1 score, prediction >=0.3 -> negative
-XGBClassifier
-DecisionTreeClassifier



MODEL COMPARISION
-Form of a dataframe
-Comparision graphs
-Result generation in a .csv file