Summary

In this segment, we will summarise all the concepts that you learnt in this session on multi-classification. Following are the topics/concepts that you learnt to understand and solve multi-classification problems:

• We analysed different business data sets that fall under the multi-classification category and also discussed the major properties in its target class column.
• We cannot apply an LR model directly to classify data samples having more than two class variables.
• The major differences between a binary problem and a multi-classification one pertain to the following:
  • Number of class variables, 2 for binary datasets and more than 2 for multi class datasets.
  • Algorithms used in each scenario. In the case of binary classification we use a machine learning model directly whereas in multi classification we use various classifier methods in which we apply machine learning models.
  • Number of models built (varies for multi-classification, unlike in the case of binary classification, where it is always one)
• The differences were explained through a business example and by plotting various discrete graphs for each class variable. The major point noted was that in a multi-classification problem, more than one LR model is trained, i.e., one model for each discrete graph.
• The different techniques used to solve multi-classification problems are as follows:
  • One vs One
  •  One vs Rest
• We discussed the One vs One classifier in detail. It consists of three steps:
  • Subset creation
  • Model building
  • Model prediction (maximum vote count method)
• We also discussed the One vs Rest classifier in detail. It consists of three steps:
  • Data set creation
  • Model building
  • Model prediction (maximum probability method)
• The major differences between One vs One and One vs Rest classifiers pertain to the following:
  • Number of data sets formed
  • Number of models created and trained
  • Method of predicting the test samples
• You also learnt about the Python implementation of both One vs One and One vs Rest classifiers and the various functions involved in the libraries to analyse the model parameters and classification results.

In the next segment, we have provided practice questions. Answer them based on the concepts that you learnt in the session.

Report an error