seolhalee / view-anlysis-for-purpose-trips

Augmenting the Visaul Experience on GPS Tracks using Machine-Learning Techniques

• RQ1. How can we infer the purpose of daily trip of people from their GPS location data?
• RQ2. Is there any common pattern of trips that people in Tokyo take throughout a day?
• RQ3. How different are the visual experience of the trips with different purposes?

I designed this study into two separate steps with two machine learning techniques respectively.

• First, I developed a prediction model of trip purpose using Xboosting framework using the extensive Person Trip (PT) survey data and predict the purpose of trips in the GPS data.
• Second, I adopted a computer vision technology called scene parsing, which segment and classify images in pixel-level with semantic categories of objects using Convolutional Neural Network.

Steps

Activity Classification

Markdown is a lightweight markup language based on the formatting conventions that people naturally use in email. As John Gruber writes on the Markdown site

View Component Analysis

(powerpoint slide placed here)

Result

Spatio Temporal Pattern of Purpose Trips in Tokyo

(image placed here)

Change in View Component along Different Purpose Trips

(image placed here)