Page 228 - AI Ver 3.0 class 10_Flipbook
P. 228

Prediction
                    Approval of Loan         AI Model                     Loan Not Approved
                                                                                                             Not
                                                                                  –                        Approved
                                                              Actual
                                                                            Loan Approved


                 • Accuracy: If the model predicts, that the loan is approved or not approved, and it matches the reality, then that
                means the model is accurate for that dataset.
                Understanding both error and accuracy is crucial for effectively evaluating and improving AI models.

                                                           Prediction
                   Approval of Loan         AI Model                      Loan Approved
                                                                                                          Approved
                                                                                 –
                                                             Actual
                                                                           Loan Approved

              Understanding both error and accuracy is crucial for effectively evaluating and improving AI models. From the
              above example we can say,

                 • The focus is to maximise the accuracy in the performance of the model and minimise the error.
                 • In real life scenarios, the accuracy is dependent on the data, if the dataset is not realistic, the best models
                may make mistakes. For example, in approval of loan, a model with slightly lower accuracy but the focus is on
                avoiding incorrectly identifying a right applicant may be preferred by the banks.
                 • Selection of the model with the balance of accuracy and error depends upon the task and its requirements by
                the model


                          Task                                                             21 st  Century   #Information Literacy
                                                                                                   #Critical Thinking
                                                                                               Skills
                Find the accuracy of the AI model mathematically.
                Calculate the accuracy of AI model which predicts the salary of employees:

                   • Using the given formulae complete the given table:
                     Error Absolute = ABS(Actual Value -Predicted Value)
                     Error Rate        =  (Error/Actual value)
                     Accuracy           = (1-Error Rate)

                     Accuracy%       = (Accuracy*100)%
                     Predicted Salary   Actual Salary  Error Absolute   Error Rate      Accuracy        Accuracy%
                    47,000             45,000          2,000           2,000/45,000  1-0.044=0.956   0.956*100=95.6%
                                                                       =0.044
                    56,000             56,500          500             500/56,500    1-0.008=0.992   0.992*100=99.2%
                                                                       =0.008
                    45,000             45,500          500

                    38,000             37,000          1,000
                    65,000             67,000          2000
                   • Accuracy of the AI model is the mean accuracy of all five samples.
                                                                       (Sum of Accuracy%)
                                             Accuracy of the AI model =
                                                                     Total number of samples


                    226     Touchpad Artificial Intelligence (Ver. 3.0)-X
   223   224   225   226   227   228   229   230   231   232   233