Final-Project-551

Description of the dataset

There are seven csv files in the dataset. Each file contains sensor data from an individual bird with the following columns.

• FID - this is just an identifier that associates each row back to the unsampled data

• tag - label for each bird

• time - time stamp for when the data were collected

• depth - depth (m)

• X - raw acceleration in the x-axis (g)

• Y - raw acceleration in the y-axis (g)

• Z - raw acceleration in the z-axis (g)

• staticX - mean of the x-axis calculated over a 2 sec window, a measure of average position

• staticY - mean of the y-axis calculated over a 2 sec window, a measure of average position

• staticZ - mean of the z-axis calculated over a 2 sec window, a measure of average position

• pitch - calculated from the static axis above, this measures the overall posture of the animal from 90 deg to -90 deg, the data have been calibrated so pitch should be close to 0 during flight

• dynamicX - dynamic movement in the x-axis based on a 2 sec window

• dynamicY - dynamic movement in the y-axis based on a 2 sec window

• dynamicZ - dynamic movement in the z-axis based on a 2 sec window

• ODBA - a composite measurement of dynamic movement in all 3 axes

• ground.speed - this is from the GPS, and should be ignored because it is inaccurate

• Temperature - this is from the GPS and may help in confirming behaviour classification

• Activity - this is from the GPS and should be ignored, because it can mean the bird wasn't moving or that the GPS could not obtain a signal

• WBF - the peak frequency of movement from the Z axis over a 10 sec window, this is useful for distinguishing flight and also potentially swimming underwater during a dive

• meanPitch240 - pitch averaged over 4 min window, this can help to distinguish when a bird is a the colony (higher pitch) vs on the water (lower pitch)

• sdODBA240 - standard deviation in the ODBA over a 4 min window, this can help distinguish when a bird is still (low values) vs active (high values)

• location.lon - eastings (m), from the GPS interpolated to 1 sec intervals

• location.lat - northings (m), from the GPS interpolated to 1 sec intervals

• speed - ground speed in km/hr calculated from the GPS

• behaviour - a rough classfication of behaviour based on the GPS and depth data, which could help with training the algorithm or checking your results

Instruction for running the code

SVM, Decision Trees, Logistic Regression, Naive Bayes

Require scikit-learn, numpy

Hidden Markov Model

Requires scikit-learn, numpy

Neural Network

Requires Keras with Theano backend, and Python3