| Individual mobility is an essential part of any transportation system,which is significantly influenced by the availability of public transport and private vehicle ownership and usage.Furthermore,controlling the rise in private vehicle ownership,and increasing quality and access to public transportation is an indispensable part of any sustainable transportation planning strategy.However,this strategy seems to elude most developing countries in Africa especially Ghana where the public transport is characterized by lowquality services culminating in rising private vehicle ownership.The rise in private vehicle ownership in Ghana has raised many concerns.According to the Driver and Vehicle Licensing Authority(DVLA),the total number of vehicles registered in Ghana increased over 700 % between 1996 and 2018 translating to an annual average increase of about 10.1% with about 80% of the vehicles registered been personal vehicles.Reasons for this rise have been attributed to rise in population,improved living conditions and increased private vehicle importation resulting from economic liberalisation,lack of investment in public transportation and uncontrolled urban spiral development.Although the overall vehicle per capita of the country is low and concentrated on a few individuals,its effects in traffic system of urban centres cannot be underrated.Existing models of vehicle ownership in most developing countries like Ghana are the aggregates models,which are insufficient in explaining the factors that contribute to the rise of individual vehicle ownership behaviour.Although the disaggregate models would have been a better approach than aggregates models,the latter is often adopted as a result of inadequate data on individual vehicle ownership.Therefore,this study attempts to fill this gap by developing a disaggregate choice model using multinomial logit(MNL)to analyse the factors influencing the rise of private vehicle ownership(motorcycle and car)in Ghana using Greater Tamale Area(GTA)as the case study.Also,vehicle ownership models can also be classified into statistical,and Machine Learning(ML)models based on the analytical technique adopted in the model development process with most vehicle ownership models developed been statistical models.However,with the advancement in computing power of computers and Artificial Intelligence,Machine Learning(ML)algorithms are becoming an alternative or a complement to the statistical models in modeling the transportation planning processes.Although the application of ML algorithms to the transportation planning processes;like mode choice,traffic forecasting and demand modeling have received much attention in research and abound in literature,scanty attention is paid to its application to vehicle ownership modeling especially in the context of a small to medium cities in developing countries.Therefore,this study attempts to fill this gap by modeling vehicle ownership in GTA,Ghana using nine ML algorithms.The study used Logistic Regression(LR),Stochastic Gradient Decent Classifier(SGD),Linear Support Vector Classification(Linear SVC),Decision Tree(DT),Random Forest(RF),Extremely Randomized Trees(ERT),Adaboast,Gaussian Naive Bayes(Gaussian NB)and K-Nearest Neighbors(KNN)ML classification algorithms to model the dataset.The performance of each classification algorithms was evaluated using accuracy,precision,recall,ROC,and Cohen-Kappa static using Scikit-Learn,an efficient and simple tool for data analysis and ML algorithms in Python.Additionally,permutation feature importance was used to determine the features that are significant in predicting vehicle ownership in GTA.We approach this by using a cross sectional survey of formal sectors workers which is the population group most likely to own vehicles and most likely to drive the increase in car ownership level in the city.The data was collected between June – August 2018 within the city with average monthly income,distance to work,provision of non-motorized infrastructure,location and sociodemographic factors like age,marital status,and household characteristics as the explanatory variables or feature variables.For the discrete choice modelling approach,the finding shows that higher average monthly income,longer traveling distance to work,increasing age,been married,been male,positively corelates with the likelihood of owning both car and motorcycle.Been a senior rank worker,higher academic qualification,higher household size and a higher number of children in the household positively corelates with a greater likelihood of owning a car whiles negatively corelates with a lower likelihood of owning a motorcycle.Living within 2km from the CBD is positively related to a higher likelihood of owning motorcycles,and the provision of adequate pedestrian and cycling lanes had a positive influence on the ownership of vehicles in GTA indicating that the provision of the non-motorized infrastructure does not discourage private vehicle ownership in GTA.Regarding the ML technique,the results show that Linear SVC is the best performing classifier in terms of accuracy of predictions whiles Linear SVC,RF,ERT,Gaussian NB and KNN are the best classifiers base on ROC.The best classifier with regards to the Cohenkappa static is Linear SVC.Therefore,based on accuracy of predictions,ROC and Cohenkappa static,Linear SVC is the best classifier for the dataset.For predictions based on class performance,KNN performs well for no-vehicle class whiles Linear SVC and Gaussian NB performs well for motorcycle ownership.For the car ownership category,Linear SVC and LR performs well in comparison with the other classifiers.The results also indicated that individual mode choice(car or motorcycle),average monthly income,average travel distance to workplace,average monthly expenditure on transport,duration of travel to workplace,been a senior rank worker,age,household size and marital status were significant in predicting vehicle ownership for most of the classifiers.However,provision of non-motorized infrastructure,living within 2km from the CBD,and willingness to switch from current transport mode to work to Metro Bus Service if available were less significant in predicting vehicle ownership for all the classifiers.The results also demonstrated that ML can also be applied to modeling and forecasting of small to medium cities in developing countries like Ghana. |
PDF Full Text Request