Statistics-based Business Insights

  1. From all available cars in the datase, the top 25% most efficient cars travel between 29 and 46.6 miles per gallon.
  2. Purchasing top 25% of most efficient cars will help us remain ahead of industry average, which is currently at 23.5.
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    9.00   17.00   22.75   23.45   29.00   46.60

Plot-based business insights

  1. The base-level MPG standard has been increasing between 1970-1982
  2. Between 1970-1973, the average MPG for cars was 18.5
  3. Between 1974-1977, the average MPG for cars was 22; a 19% increase
  4. Between 1978-1981, the average MPG for cars was 28; a 27% increase
  5. In the year 1982, the average MPG for cars was 32

Logistic regression analysis: business insights

  1. From all given variables, acceleration and horsepower are the 2 biggest influences in the likelihood of a car travelling 30+ miles per gallon.
  2. With every increase of acceleration by 1 unit, the probability of success decrease by 23.5%
  3. With every increase of horsepower by 1 unit, the probability of success decrease by 10.24%
  4. With every increase of cylinders by 1 unit, the probability of success decrease by 5.56%
  5. With every increase of weight by 1 unit, the probability of success decrease by 0.09%
## 
## Call:
## glm(formula = good_bad ~ cylinders + horsepower + weight + acceleration, 
##     family = "binomial", data = mydf)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -2.2241  -0.4480  -0.0838  -0.0006   3.3354  
## 
## Coefficients:
##                Estimate Std. Error z value Pr(>|z|)    
## (Intercept)  14.7300399  2.5194368   5.847 5.02e-09 ***
## cylinders    -0.0572518  0.3564901  -0.161  0.87241    
## horsepower   -0.1080746  0.0228523  -4.729 2.25e-06 ***
## weight       -0.0009030  0.0007643  -1.181  0.23742    
## acceleration -0.2676784  0.0976733  -2.741  0.00613 ** 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 422.41  on 391  degrees of freedom
## Residual deviance: 226.93  on 387  degrees of freedom
## AIC: 236.93
## 
## Number of Fisher Scoring iterations: 8

Tree analysis: business insights

  1. In the dataset, 302 cars travel up to 29 miles per gallown, and 90 cars travel 30 or more miles per gallon
  2. For cars that have horsepower of 76 or less, 71 of those travel 30+ miles per gallon
  3. For cars that have horsepower of 77 or more, 19 of those travel 30+ miles per gallon
  4. For cars that have horsepower of 86 or more, 8 of those travel 30+ miles per gallon

Model comparison diagram

After comparing the Logistic Regression model against the Tree model, the data concludes that the Tree model is more accurate. Please use Tree model to evaluate business insights of the Cars dataset.