How does taxi driver behavior impact their profits?-discerning the real driving from large scale GPS traces

Name: Thananut Phiboonbanakit

keyword: Recommendation system

LCSH: Taxicabs

; Taxi profit

LCSH: Data mining

Abstract: With a trend towards the use of large scale vehicle probe data, the entire urban scale analysis is become possible to suggest useful information for taxi drivers and passengers. In our research we used Rama, I as representative of Phra Na Korn Side and BTS Wong Wien Yai as the representative of Thonburi side to make comparison. We also introduce taxi trip assessment model to make evaluation on taxi trips This study, first, we have data exploration process to find trend and pattern of our obtained data. Then we make use the advantage of geospatial tools such intersection and buffering to divide and clustering it into areas. Then we calculate profit using taxi trajectory calculation algorithm by obtained fare rate and reconstructed trips. This process is running on the most advance tool of computing for the large-scale data such as Hadoop. Second, the data were analyzed by using mathematic model to understand distance profit, total profit, total net profit in timely basis. The mathematic model is calculated taxi consumption cost, net profit, and energy resource cost. The taxi consumption cost is the cost occur from the normal drive and the consumption of the engine. The net profit is the cost where the total cost has been deducting by the consumption cost. Third, we calculate probability of taxicab to observes chance that taxi drivers will get customer in the specific area. This could recommend taxi driver to make decision and plan the situation if they willing to come to this area or not. Forth, we analyze on the taxi driver working hour and expense when they exit from the gas station. This would give the ground truth of actual cost that taxi driver will expense when they enter the gas station each day. iii Finally, we built a model for fit and evaluate a result from our calculation model with real taxi trip which we collect from real driving of taxi drivers (Ground truth). We build to models. One is evaluate the accuracy of predict net profit and other model used to evaluate the classification of the worthiness level on the taxi trips. This step will be the final step to evaluate our work so far if it is conducting the correct way or not. The result indicated that the pickup rate of taxi in this area is usually peak per area of interest and operation hours. The increased profits were mainly based on the distance and time of each trip. The trip run in shorter distance more frequency (9 times compare to one long distance) give high profit than the long distance. We also discover that most of taxi get customer from detour (On driving) in Rama I area. The result is varied on environment and area that we selected. Forth, taxi driver working hour is at least 14 hours per day from the data which we have been collated. Finally, Random Forest regression is the easiest for model tuning and configuration over Random Forest and Decision Tree. It gives the prediction error only 10 Baht error from the actual net profit. We used the average profit from our ground truth it obtained over 90 percent accuracy. For the worthiness level classification, it took about 89 percent accuracy. In the evaluation process, we used Root Mean Squared and R-Squared to evaluated profit prediction model. Also, the worthiness classification model, we used confusion matrix to evaluate our model. These results uncover taxi driving behavior in Bangkok and yields great benefit for both taxi drivers and passengers. We also come up with solution to make taxi driver earn more profit and reduce decline service to customer by adding fare-rate to regular fare to make drivers earn more profit. In distance, less than 10 kilometers will add fare-rate 15 baht, 10 to 20 kilometers will add 30 baht and more than 20 kilometers will add 45 baht

Thammasat University. Thammasat University Library

Address: BANGKOK

Email: preserv@tu.ac.th

Name: Teerayut Horanont

Role: advisor

Created: 2016

Modified: 2021-06-01

Issued: 2021-06-01

วิทยานิพนธ์/Thesis

application/pdf

DOI: 10.14457/TU.the.2016.522

eng

DegreeName: Master of Science

Level: Master's Degree

Descipline: Engineering and Technology

Grantor: Thammasat University

©copyrights Thammasat University

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	2571thananut.pdf	2.59 MB	1	2023-05-12 18:27:25

ใช้เวลา

0.02624 วินาที