Page 299 - Ai Book - 10

7 7 Evaluation
Evaluation

u Model Evaluation u Evaluation Matrix for AI Model
u Confusion Matrix

MODEL EVALUATION

The final stage in the AI project cycle is evaluation of the trained model. Model evaluation is a crucial part of the
AI project cycle and it helps us identify performance and possible changes to improve it. Evaluation refers to the
process of understanding the reliability of any AI model by feeding a test dataset that has never been used for
training. At the time of evaluation, we should never consider those datasets which are used for training as our
model will always predict the correct label or output for such datasets. Let’s explore various terminologies used
in Model evaluation with the help of a scenario.
Scenario: Suppose you have given a task to deploy an AI based prediction model in the area which is prone to
flood. Now, the objective of the model is to assess the area and predict whether the flood has occurred or not.
In this case, two terms or conditions are used to determine the efficiency of the model. These terms are:

u Prediction: Prediction is the output generated by the machine.
u Reality: Reality is the real condition of the area when the prediction has been made.
Now, we can make various combinations with these conditions like:
Case I: Does the flood occur there?

Prediction: Yes True Positive Reality: Yes

173
173