| CAR DESIGN ANALYSIS | | | | SUBMITTED BY :FT12206 BALAJI MADHUSUDHAN |

|

Car design Analysis
Sum of Squares comparisons: Technique | Total sum of squares | MLR | 1989.69 | ANN | 1716.93 | GA | 496.5 |

Results: * As can be seen from the above table, Multiple linear regression technique (MLR) provides the model with highest “Total Sum of Squares” followed by the Artificial Neural Network (ANN) anby Genetic Algorithm (GA). * The three models give different results because because MLR assumes a linear relationship between the dependent and independent variables while both ANN and GA don’t have any such assumptions and they form models based on the pattern observed from the data. * If we compare ANN with GA we can see that GA has a better model in terms of “Total sum of squares”. This could be due to the absence of prior solution for ANN to form a model or due to presence of noise in the data. * We were estimating how much of each of the independent variables would result in dependent variable. Since we were not performing classification but were estimating the dependent variable, GA was the appropriate technique under the context. * Hence we observe an enormous difference in results between the three when compared the “Total sum of squares”.
Situational based technique: * ANN models are more accurate than any other technique but the disadvantage of ANN is the lack of explainability. Also, ANN is used generally when we have to come up with a classification model. * GA model is less accurate when compared to ANN but has higher expainability than ANN. * ANN is generally used to answer the “Where” type of problems while GA generally answers “How” type of problems because the explainablity is more in the GA. * Thus whenever the problem demands for more explainability than