Question-26: Can you tell me; how would you productionize the Spam Detection Machine Learning Algorithm?

Answer: Let’s say, we have 100,000 of historical emails, which are already labeled like Spam or not Spam. We define our Machine Learning algorithm and train it with these all historical emails, until it can predict all the emails correctly. Once we get almost 100% accuracy, we can put this Machine Learning Model or algorithm in production to categorize all new future emails, whether those are spam or not spam.

Question-27: What is multi-class classification?

Answer: As name suggests, when we classify items in more than two categories, that is called Multiclass Classification.

Question-28: Can you few more examples of classification Machine Learning?

Answer: Yes, below are few more examples to clear our understanding about classifications.

• Disease Prediction: Based on historical information from millions of patient’s responses a model can be created and trained to predict the disease for the new patients. We are also seeing now a day’s online applications are available they ask few questions and based on that they predict possible disease.
• Customer Churn: When customer can stop using services, like unsubscribing from a subscription. We can use all our existing and past customer. Based on their behavior we can create a binary classification model. Whether or not the subscriber would continue our services or not.
• Prediction on Buy or not Buy online: You are having an eCommerce website and need to find whether customer will buy a product or not on the website. To do that, you can use previous history of the data and train your model accordingly for instance browsing patterns, customer location etc. And your model can predict whether visitor is a potential customer or not. And based on that you can pop some offers for him instantly.

Question-29: What is the regression technique in Machine Learning?

Answer: As with the classification Machine Learning technique we tried to predict some discrete values like yes or no. Buy or not buy. Spam or not spam etc. However, in regression we would have to predict some continuous variables also you can say real number. So rather than you predict category, your model will be predicting number on some line. However, all other stuff is same as other classification Machine Learning model. Hence, we can say regression is also a Supervised Machine Learning.

Question-30: Can you give some examples of regression (Supervised Machine Learning)?

Answer: Yes, below are the few examples of regressions

• Online Sales Prediction: Use last 10 years sales data and create and train model using this data to predict the next month sales.
• Height Prediction: Based on the historical data which include Parent’s height, location e.g. Hilly area, forest area, plain area etc. You want to predict the height of the children based on their age and what maximum possible height they can reach.
• Prediction of viewers on day one: You want to have a model based on the character’s in the movie, story of the movie, current trending topic, epidemic etc. And model should predict the number of possible viewers on day1, day2, day3 of the movie release.