Table 1.
Describing features types, divided for category which have been collected in our data collection process.
Table 2.
Describing features types, divided for category which have been collected through the distribution of our survey.
Fig 1.
Regional features average Pearson correlation coefficient with the target variable (regional percentage of EVs).
The average is computed as the mean value of yearly correlations. Values are included in the range [–1,1].
Fig 2.
Scatter plot matrix with the target variable (regional percentage of EVs) of the following features: GDP Per Capita (PIL), Polluting Potential, PM10 Exceedances, Number of Photovoltaic Panels and Number of Charging Points.
Fig 3.
Workflow of the Training and testing procedure of the model for the regional and provincial analyses.
Table 3.
Performance for the four different applied models in regional analysis: Linear Regression (LR), Ridge Regression (RR), Decision Tree Regressor (DT), Extreme Gradient Boosting Regressor (XGBR).
The last row shows the metric’s average value for the four years.
Fig 4.
Scatter plots matrix for regional models’ performance.
For each scatter plot, the true values are reported on the x-axis, while the predicted ones are reported on the y-axis. The rows of the matrix represent the years, while the columns represent the implemented model: Linear Regression (LR), Ridge Regression (RR), Decision Tree Regressor (DT), Extreme Gradient Boosting Regressor (XGBR).
Fig 5.
Provincial features average correlation with the target variable (provincial percentage of EVs).
The average is computed as the mean value of yearly correlations. Values are included in the range [–1,1].
Fig 6.
Scatter plots matrix with the target variable (regional percentage of EVs) of the following features: Large Companies, Micro Enterprises, Polluting Potential, PM10 Exceedances, Number of Photovoltaic Panels and Number of Charging Points.
Fig 7.
Regional selected features with correlation equal or higher than 0.3 in absolute value.
Values are included in the range [–1,1].
Table 4.
Performance for the four different applied models in provincial analysis: Linear Regression (LR), Ridge Regression (RR), Decision Tree Regressor (DT), Extreme Gradient Boosting Regressor (XGBR).
The last row shows the metric’s average value for the four years.
Fig 8.
Scatter plot matrix for provincial models’ performance.
Fig 9.
Respondents’ opinion towards photovoltaic panels.
Starting from the left, it shows how many respondents own at least a Photovoltaic Panel, their Buying Propension with and without it.
Fig 10.
The percentage of users owning a Garage and their propensity to Install a private charging point.
Fig 11.
Respondents’ opinion towards the main factors influencing the spread of electric cars.
They had to assign a weight from 1 to 5 to each of the factors.
Fig 12.
EVs owners’ opinion towards the main factors influencing the spread of EVs.
Fig 13.
Current and future buying propensity of respondents.
Table 5.
Weighted average scores for current propensity predictions over the 3 run of the CV procedure.
Table 6.
Weighted average scores for future propensity predictions.
Fig 14.
The workflow followed for the research.
Starting from the regional analysis (red), the focus of the research was restricted to the province (blue), up to analyzing a sample of the population through the administration of a survey (black).