CaGeo: CAnonical Geospatial features

In this repository we want to collect and implement methods to extract from a set of raw GPS coordinates features that enrich the dataset, but which, simultaneously also allow dimensionality reduction.

Features:

Point Features

• Speed: AVG, STD, MIN, MAX
• Acceleration: AVG, STD, MIN, MAX
• angle/direction: atan2 == Bearing: angle between the magnetic north and an object ??

Aggregate features

• Turning Angle: AVG, STD, MIN, MAX. 𝐻𝐶𝑅 = |𝑃𝑐|/𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 Pc is the collection of gps points at which a user changes his/her heading direction exceeding a certain threshold (Hc), and |𝑃𝑐 | represents the number of elements in Pc
• Traveled Distance: SUM
• Stop Rate: 𝑆𝑅 = |𝑃𝑠|/𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 Ps is the collection of point with velocity smaller than a certain threshold
• Velocity Change Rate: foreach point 𝑝1. 𝑉𝑅𝑎𝑡𝑒 = |𝑉2 − 𝑉1|/𝑉1; then 𝑉𝐶𝑅 = |𝑃𝑣|/𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 where 𝑃𝑣 ={𝑝𝑖|𝑝𝑖 ∈ 𝑃, 𝑝𝑖 . 𝑉𝑅𝑎𝑡𝑒 > 𝑉𝑟 }
• FFT?
• duration of movement?
• traveled path?
• displacement?
• Bearing rate: B_rate(i+1) = (Bi+1 − Bi)/∆t
• Rate of bearing rate: Br_rate(i+1) = (Brate(i+1) − Brate(i))/∆t

Derivate features

• sinuosity ?
• distance from POI

References:

• A survey and comparison of trajectory classification methods
• Understanding mobility based on GPS data
• Revealing the physics of movement: comparing the similarity of movement characteristics of different types of moving objects
• Predicting Transportation Modes of GPS Trajectories using Feature Engineering and Noise Removal
• Determination transportation mode on mobile phones

Note

Per le distanze vedere Intelligent Trajectory Classification for Improved Movement Prediction

In "Identifying Different Transportation Modes from Trajectory Data Using Tree-Based Ensamble Classifier": ci sono varie misure globali