Classification Models Flashcards
In the context of ML, what is the definition of classification?
A type of supervised ML that uses labels to represent a class or category of data
How do algorithms for classification work?
Algorithms predict the class based on probability not numeric values
What are two types of classification algorithms?
Binary and multiclass
What is binary classification?
Using one feature to determine if an object is or isn’t in a certain class. This is also used to say if something is true/false or 1/0
How does multiclass classification work?
An algorithm classifies or predicts the class of an object from either among multiple classes (2 or more) or if it fits multiple categories (like book and movie genres)
What is logistic regression?
A type of classification for Boolean (true/false, positive/negative) values.
What is the difference between logistic and linear regression?
The shape of the curve: Linear is a straight line and logistic is s-shaped. (Optional) This happens happens because logistic regression produces values of 1 or 0 and anything in between is considered a probability (ex: probability of an earthquake could be predicted as 0.3 or 30%)
True/False Logistic regression can be applied to situations with 3 or more outcomes
True: you would have to use multinomial logistic regression for it.
How are classification models assessed?
By their output probabilities or the final labels used like saying there is a 20% chance of rain or using the category of ‘rain’, ‘no rain’.
What is MSE?
Mean squared error
How is MSE calculated?
By comparing the model’s prediction with the actual label, square the difference, and taking the average of the result.
What is the function of log loss and MSE?
To calculate cost/loss or how badly a model performs
What is the difference between log loss and MSE?
Log loss is more demanding than MSE having higher cost/loss for wrong predictions. This is used to make a model learn faster.
What is the limitation of single cost functions?
It doesn’t tell you what kind of mistakes the model is making, just that its making them.
