Page 227 - AI Ver 3.0 class 10_Flipbook
P. 227

Reboot


                       1.  What are the types of evaluation techniques?


                       2.  What characterises a model experiencing underfitting?






                         Accuracy and Error


                 Let’s consider a scenario where, Aman and Priya went grocery shopping to buy a bag of 5 kg rice. Aman got `450
                 and Priya got `500. The actual price of the rice was `550.
                    • Who is more accurate? Aman or Priya?
                   Priya was more accurate, as her estimate is closer to the actual price.
                    • How much is the error for both Aman and Priya in estimating the price of a bag of rice?
                   Error is the difference between the measured value and true value.
                   Aman’s Error is `100 (550-400)
                   Priya’s Error is `50 (550-500)
                   Here, Priya is more accurate as her error is less than Aman’s error.
                 The term Accuracy is defined as the evaluation metric that measures the total number of predictions that are
                 correct by the model. It means how close the prediction is to the true value. The accuracy of the model and the
                 performance of the model is directly proportional, which means better the performance of the model,  leads to
                 higher accuracy in predictions.
                 The term Error means the action that is inaccurate or wrong. It refers to the difference between a model’s prediction
                 and the actual outcome. It quantifies how often the model makes mistakes. Based on the error, we choose the
                 machine learning model that has the best performance for a specific dataset. Low error in a model 's performance
                 signifies precise and reliable predictions.
                                                             Based on the present error, the AI model
                                                         parameters are fine tuned to reduce further error


                                                                        Predicted Value

                              Input Data            AI Model                 less           =       Error
                                                                         Actual Value


                 Note, high accuracy in a model indicates better model performance, but it
                 may not reflect the true scenario, especially with an imbalance dataset.

                 For  example,  you’re  training  a  model  to  predict  the  approval  of  loan
                 (Classification Task)
                    • Error: If the model predicts that the loan will not be approved but actually
                   it’s approved by the bank, that’s an error. The error is the measurement
                   of the difference between the prediction and the actual outcome.



                                                                                           Evaluating Models    225
   222   223   224   225   226   227   228   229   230   231   232