Page 135 - Ai_V1.0_Class9
P. 135

• Evaluation ensures that the model is operating correctly and optimally.
                    • Evaluation is an initiative to understand how well it achieves its goals.

                    • Evaluations help to determine what works well and what could be improved in a program.


                 Model Evaluation Terminologies
                 Evaluation of an AI model can be done using various terminologies. Let us try to understand them with the help
                 of a scenario.
                 Scenario: An AI-based prediction model is deployed in schools. The model is supposed to predict whether the
                 students of grade 10 will be able to get a “Perfect Score“ in the board exams this coming year or not. The model
                 will check for whether the students will be able to score full marks in board exams in the coming year or not.
                 There are two important parameters that are used for the Evaluation of a model. These are:
                 Prediction: It is the output given by the AI model using a Machine Learning algorithm.
                 Reality: It is the real scenario of the situation for which the prediction has been made.
                 Let’s look at the various combinations that can be considered for the above scenario.


                 Case 1: Is There a Perfect Score?
                 Due to coaching institutes and a lot of easy access to online
                 resources, it is not clear whether this would work in their
                 favour or not. Guidance and coaching are so readily available
                 to all that most of them are prepared well for the exams, but
                 it can also lead to tougher papers to test them well. This
                 makes getting a perfect score unpredictable. We need an
                 AI model that can  predict  whether  the  student  can  get  a
                 Perfect Score or not in the board exams so that the students   Predicion: Yes                Reality: Yes
                 can timely plan their preparation and schedule to study as                    True Positive
                 per the exam schedule.
                 In the above picture, we show the possibility of students scoring full marks in board exams for grade 10. The model
                 predicts a Yes, which means the student will get a Perfect Score in board exams to be conducted. The prediction
                 matches with the reality: Yes, therefore, this condition is called True Positive.


                                                                Case 2: Is There a Perfect Score?
                                                                There  is no  Perfect  Score  as the  Examination  Paper  was
                                                                difficult, hence the Reality is No. In this case, the machine too
                                                                has predicted it correctly as a No. Therefore, this condition is
                                                                termed as True Negative.

                    Predicion: No                   Reality: No
                                   True Negative


                 Case 3: Is There a Perfect Score?


                 Here, the reality is that  there are no Perfect Scores due
                 to  difficult  papers.  However,  the  machine  has  incorrectly
                 predicted that there will be a Perfect Score for the students   Predicion: Yes                Reality: No
                 of grade 10. This case is termed as False Positive.                           False Positive


                                                                         AI Reflection, Project Cycle and Ethics  133
   130   131   132   133   134   135   136   137   138   139   140