Browsing by Subject "Predictive models"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Designing an Autonomous Service to Cover Transit’s Last Mile in Low-Density Areas
    (Minnesota Department of Transportation, 2024-03) Khani, Alireza; Aalipour, Ali; Kumar, Pramesh
    Public transportation provides a safe, convenient, affordable, and environmentally friendly mobility service. However, due to its fixed routes and limited network coverage, it is sometimes difficult or impossible for passengers to walk from a transit stop to their destination. This inaccessibility problem is also known as the "transit last-mile connectivity problem." Such a lack of connectivity forces travelers to drive, thereby increasing vehicle miles traveled (VMT) on roads. The autonomous mobility-on-demand (AMoD) service, with characteristics such as quick fleet repositioning and demand responsiveness, as well as lower operational cost due to the elimination of operators' wages, has the potential to provide last-mile coverage where fixed-route transit can only provide limited service. This study presents research on designing an AMoD service to solve the transit last-mile problem in Greater Minnesota. After selection of the Miller Hill Mall (MMH) area in Duluth, MN, as the case study site, analysis on local transit services and demand data show that passengers may have to spend significant time walking and cross multiple streets to access stores from transit stops. To address this issue, an AMoD system for last-mile service was designed and integrated with the fixed route transit service. Novel mathematical models and AMoD control algorithms were developed, and simulation experiments were conducted for evaluation of the AMoD service. Simulation results showed that the AMoD service can improve transit quality of service and attract more riders to use transit to the MHM area, and therefore reduce the VMT in the region. These findings were consistent with the literature in that mode choice and first-/last-mile access were highly interdependent and AMoD can improve transit quality of service and reduce VMT. Research on riders' perception of AMoD service and field testing of the AMoD system using the developed models and algorithms are recommended to help agencies prepare for application of AMoD system in the region.
  • Thumbnail Image
    Item
    Statin-Associated Adverse Events Prediction and Drug-Drug Interactions for Cardiovascular Disease Patients from Retrospective Claims Data
    (2019-08) Wang, Jin
    Statins are commonly used to lower cholesterol levels for cardiovascular disease (CVD) patients in the primary and secondary prevention of acute events. 26% of American adults over age 40 used statins in 2012 and an estimated 26.4 million U.S. adults could benefit from statin use. Although statins are generally well tolerated and show a relatively good safety profile, concerns have been raised regarding statin associated adverse events (AEs) especially muscle related events, leading to medication non-adherence and discontinuation. Besides, AEs are often caused by potential drug-drug interactions (DDIs) which are responsible for up to 2.8% of hospital admissions[58]. Among CVD patients, combination therapy of statins and other medications is highly likely, which results in altered absorption, distribution, metabolism, or excretion of statins and thus causes adverse events. Traditional AE management approaches may include a statin therapy holiday, lower statin dosage, an alternative statin agent, or non-statin cholesterol-lowering therapy. Currently, there are no tools to effectively predict and reduce the risk of AEs prior to statin therapy initiation. In addition, no population-based studies have focused on a specific statin and a specific interacting drug and differentiated their risks among different study time periods. In this study, we investigated the effect of combination therapy of simvastatin and several pre-defined high risk interacting drugs, which belong to cytochrome P450 (CYP) 3A4 and/or organic anion transporting polypeptide (OATP) inhibitors, in CVD patients who used simvastatin for secondary prevention. This could provide some evidence and recommendations for selected interacting drugs used in CVD patients. In addition, we aimed to build a model to predict statin-associated AEs that may reduce the risk of statin associated adverse events and the rate of statin therapy cessation. Several machine learning methods were applied, such as generalized linear model (GLM), support vector machine (SVM), decision tree, random forest, and artificial neural network (ANN). Models were developed and compared for their performance. The best model was selected based on the best performance.