Modify and add to the Python code using deep learning and related to the analysis of tweets
Budget: $30 – $250 USD
I have python code that is executing on google colab
Its purpose is to classify tweets as either fake or real
This code cleans datasets (which is text for tweets) and preprocess it
and then tokenizing, embedding,
And then balancing the data using one of four balancers (random over sampler, SMOTE, random under sampler, near miss).
And then the model implementation, which uses the LSTM algorithm
Then make train, test , measure confusion matrix
The following is required in several stages:
- Make a combination between two balancers (ex: random under sampler + random over sampler )
So that the proportion of the minor increases and the proportion of the major decreases without being equal
(needed to be done within 48 hours max)
- The current code works only on the "text" column.. It is required to work on the rest of the columns (username, ID number, meta text, number of followers, number of favorites, number of retweets, number of friends, etc.) and the results are in acceptable proportions (ranging between 60 and 95%)
(needed to be done in 72 hours max)
+
need to clarify the changes of the code
Its purpose is to classify tweets as either fake or real
This code cleans datasets (which is text for tweets) and preprocess it
and then tokenizing, embedding,
And then balancing the data using one of four balancers (random over sampler, SMOTE, random under sampler, near miss).
And then the model implementation, which uses the LSTM algorithm
Then make train, test , measure confusion matrix
The following is required in several stages:
- Make a combination between two balancers (ex: random under sampler + random over sampler )
So that the proportion of the minor increases and the proportion of the major decreases without being equal
(needed to be done within 48 hours max)
- The current code works only on the "text" column.. It is required to work on the rest of the columns (username, ID number, meta text, number of followers, number of favorites, number of retweets, number of friends, etc.) and the results are in acceptable proportions (ranging between 60 and 95%)
(needed to be done in 72 hours max)
+
need to clarify the changes of the code