Can you build a model to predict toxic comments?

Another post starts with you beautiful people!
Hope you have learnt something new and very powerful machine learning model from my previous post- How to use LightGBM?
Till now you must have an idea that there is no any area left that a machine learning model cannot be applied; yes it's everywhere!
Continuing our journey today we will learn how to deal a problem which consists texts/sentences as feature. Examples of such kind of problems you see in internet sites, emails, posts , social media etc. Data Scientists sitting in industry giants like Quora, Twitter, Facebook, Google are working very smartly to build machine learning models to classify texts/sentences/words.
Today we are going to do the same and believe me friends once you do some hand on, you will be also in the same hat.

Challenge Link: jigsaw-toxic-comment-classification-challenge
Problem: We’re challenged to build a multi-headed model that’s capable of detecting different types of toxicity like threats, obscenity, insults, and identity-based hate.
Solution Format: We need to create a model which predicts a probability of each type of toxicity for each comment.

Let's first start our analysis by digging into the given datasets:-
#loading the datasets from kaggle

#checking the shape of the datasets

#Lets check the class imbalance in the train dataset

#check and fill NA

#store the target in a variable

Next, we will work on the cleaning of texts. We will write a function where we will remove some clean text using regular expression package-

Let's apply above function to both train and test datasets-

Next, we will split the train and test datasets-

Now we will transforms text to feature vectors that can be used as input to estimator and for this purpose we will use TfidfVectorizer-

Next, we will learn the vocabulary from the feature vectors , then use it to create a document-term matrix-

We will transform the test data into a document-term matrix-

Now it's time to import and instantiate the Logistic Regression model-

Since the target variable is more than one, we will use for loop and apply our model and then we will see the prediction and accuracy of our model-

From above technique I got below accuracy-

And last we submit our prediction in Kaggle as below-

If you want to see the actual code , you can see it from here- Kaggle Toxic Classify

The above model is just for your reference so that atleast you can start your text classification journey in Machine Learning. So please download my code, improve the logic, apply different models, try to improve the score and share your learning.

Meanwhile Friends! Go chase your dreams, have an awesome day, make every second count and see you later in my next post.