الفهرس 
BACKGROUNDOnly 14 pages are availabe for public view
 |  | from | 120 |  |  |
| | | from | 120 | | |
Abstract
Internet of Thingsis concluded in reading and scanning On-Board Diagnostics (OBD) data in vehicles. An OBD Adapter is connected to the Data Link Connectorin the vehicle to scan and read the Electronic Control Unit via a Bluetooth circuit. The Electronic Control Unit links different modules in the vehicles. We send the scanned OBD data to a remote server using an android application to be stored. The scanned OBD data includes some OBD features that affect the instant Fuel Consumption (FC) values.
The proposed model aimed to predict FC using Support Vector Machine (SVM) as a Machine Learning (ML) algorithm. We used a SVM algorithm to predict FC and enhance the capability of decreasing the FC levels in vehicles. We used a historical FC Dataset to feed and train our model including 8262 records of FC data.
Our prediction model depended on VS, Mass Air Flow (MAF), Revolutions Per Minute (RPM), and Throttle Position Sensor (TPS) features in our legacy FC Dataset. In addition to the collected OBD dataset which is gathered using Internet of Things technique. The accuracy of our model reached 0.97 during the testing phase. This result refers to a high accuracy value for our SVM model, which is measured by the coefficient of determination metric, R-Squared/R2.
Our observations of the SVM indicate that our proposed model looks more accurate when compared with other related studies. We had achieved higher results than other related works depending on the R2 metric, using the same algorithm. SVM can now be considered a vital algorithm for FC prediction purposes. Manufacturers must look after SVM and consider it during their observations when they discuss or implement FC prediction purposes.