How do you choose a classification algorithm for a specific machine learning problem ?

First of all I am a newbie with machine learning just trying my hands here and there . I have a specific problem in hand i.e If I have contact access of one person A , I have to classify the relation of the contact to A . Suppose I have contact list like this

Abhishek +91 898989XXXX
Dad +91 898989XXXX
Puja Aunty +91 898989XXXX
Sid Bro +91 898989XXXX
Mom +91 898989XXXX
Jeet Ganguly Sir +91 898989XXXX

I want to predict the relation from contact name like

Abhishek - Friend
Dad - Parent
Puja Aunty- Parent
Sid Bro- Friend
Mom -Parent
Jeet Ganguly Sir - Teacher
Abhishek's Dad - Parent

This problem is for a parent teacher student communication app . So how would you predict this ? I have previously solved poblems like

twitter sentiment analysis using textblob
gender classifier from height weight foot size using svm
city name generator using LSTM

The problem in hand is which classification algorithim I should use for this text analysis prediction and how should one should choose the perfect algorithm .

Responses(1)

Cross validation is one initially you should look read below link en.wikipedia.org/wiki/Cross-validation_%28statist… http://en.wikipedia.org/wiki/Model_selection few other things you should consider

Training needs to be done incrementally or batch? if incrementally then Bayesian works best.Neural network and SVM works well with batch and when you have enough data 2.how fast you want the result? SVM is very fast for classification problem because they only need to figure it out which side of plane you data point is placed
does your data composed of categorical, numerical or both. 4.Do you need to explain in details the algorithms to you manager, audience then Choose decision tree, random forest. Neural nets are hard to explain because mostly the are black box.

Similarly i can write hundreds of reason. I would suggest go to Kaggle and looks at all kind of classification and check what are algorithms people are applying. Its very hard to write each and every case here. this is the skill every data scientist need to learn by their-self.

Thread

How do you choose a classification algorithm for a specific machine learning problem ?

Responses(1)

Recent threads

Search Hashnode

How do you choose a classification algorithm for a specific machine learning problem ?

Responses(1)

Recent threads