• Battery Consumption
• Distance calculation
• EV
• Nodes
• Road network

There are tons of motivations for people to move toward EVs. First, it is more economical to use electrical energy in comparison with gas. The second reason is that deploying EVs can significantly reduce GHG emissions and air pollution. Third, EV has far fewer moving parts than an ICE and it causes to eliminate almost all maintenance costs. EVs are silent vehicles and people prefer to drive with more tranquility mode. These reasons encourage researchers to find a way for decreasing the consumption of EVs.
The mobility data is the source data for this research work, and it has done much research works in this area. The mobility data is the movement behavior of people (GPS, Mobile data, etc), and the input data for analysis is the GPS data which is collected for a period from vehicles in Tuscany.
The data are of significant importance for companies working on allocation of electric vehicles recharging services to supply of electric vehicles charging demand as it is important to find the optimal location to install electric vehicles charging station that meets their charging needs within the all-electrical driving range of their vehicles.

This thesis tried to make a machine-learning-based prediction model with the actual vehicle’s tracking data(GPS) and almost for all the machine-learning projects it is crucial to spend much more time in the preprocessing part to analyze and create a model. This work tries to find a solution to calculate the consumption on the links of the road network which have no information about them. It means that vehicles with GPS set did not pass these links and it is important to do similar calculation also for this part of the network.

The first work after reading the data-set is cleaning some rows with an unrelated or missing value. There are some values that are not useful and should be removed or replaced with other values, for instance, there are rows with points show speed to value “0” and we have deleted them. The next step is using the map-matching technique that can change the position of the points that are not in the road network.

Here, there is a trajectory-based battery consumption simulator that gets input as GPS information and creates different trajectories with the points to make a calculation for these trajectories. This time we change it to get input, as python’s output is (CSV of segments). After map-matching in Python, there is an output file with 6 columns, Lat1, Lon1, Lat2, Lon2, Speed, Timestamp for each segment. After the simulator again in Python, we extract additional information from Tuscany Graph and add it to the CSV, which will be used for training in the ML part.
Last, but not least, in this part, the main goal of this work must be done. Making a prediction model with ML Linear Regression algorithm, this model can predict part of the road network which have no information on hand about them. Therefore this part applies the model to predict the consumption for the parts of the road network which are missed.