Mobility Prediction task using latitute and longitute values with visualization.
Problem Statement:
-
To predict the path of a new entity given a record of trajectories of people/records in terms of latitude and longitude values with timestamps.
-
To visualize the highest likely path and plot it on a map that shows the trajectory of the new entity for 40 instances/time-periods of the path. (Lower timestamp intervals of 1, 10 or 15 mins preferred)
Idea reference origin url: https://www.nytimes.com/interactive/2019/12/19/opinion/location-tracking-cell-phone.html
Dataset used (or under consideration):
People: https://www.microsoft.com/en-us/download/details.aspx?id=52367 https://archive.ics.uci.edu/ml/datasets/GPS+Trajectories
Car: https://smoosavi.org/datasets/dact https://www.microsoft.com/en-us/research/publication/t-drive-trajectory-data-sample/?from=http%3A%2F%2Fresearch.microsoft.com%2Fapps%2Fpubs%2F%3Fid%3D152883
Approach:
- To use RNN to predict the path of the object.
- To cluster the paths defining hexagonal areas. Then find the probabilities of the path crossing that area.
Conclusions:
Bi-directional LSTM does better compared to LSTM and gru in path regression
Deadline:
Friday, 23rd July, 2021 EOD/ Saturday 24th July, 2021 Afternoon
Links to Resources:
https://github.com/uber-web/kepler.gl-data https://www.quadrant.io/resources/location-data https://datarade.ai/ https://github.com/anitagraser/movingpandas https://www.kaggle.com/stoney71/new-york-city-transport-statistics?select=mta_1712.csv https://medium.com/nightingale/twelve-million-phones-one-dataset-zero-privacy-an-interview-with-the-new-york-times-stuart-a-e2988d398ba3 https://archive.ics.uci.edu/ml/datasets/GPS+Trajectories https://crawdad.org/ https://smoosavi.org/datasets/ https://smoosavi.org/datasets/dact https://data.cincinnati-oh.gov/Efficient-Service-Delivery/Vehicle-GPS-Data-Department-of-Public-Services/b56d-ydmm https://archive.ics.uci.edu/ml/datasets/Taxi+Service+Trajectory+-+Prediction+Challenge,+ECML+PKDD+2015 https://snap.stanford.edu/data/loc-brightkite.html https://snap.stanford.edu/data/loc-gowalla.html https://privamov.github.io/accio/docs/datasets.html