MedadRufus/MastersProject

Issues

• Groups the features like this

  #175 opened 3 years ago by MedadRufus
  0

• Show the geographical spread of data with a plot like this

  #181 opened 3 years ago by MedadRufus
  1

• Calculate true range of each trip

  #187 opened 3 years ago by MedadRufus
  0

• Compare predictions in region trained before vs not trained.

  #177 opened 3 years ago by MedadRufus
  0

• Convert all instances of labels to contain units.

  #184 opened 3 years ago by MedadRufus
  1

• Say that pressure is measured in a box, not exposed to wind

  #189 opened 3 years ago by MedadRufus
  0

• Show how accurate baro pressure is vs gps

  #190 opened 3 years ago by MedadRufus
  0

• Increase number of bins in pair plot

  #171 opened 3 years ago by MedadRufus
  0

• Try the hybrid SOC/no SOC model

  #180 opened 3 years ago by MedadRufus
  0

• Use Other ML models, such as Light BGM and structured regression models RNN

  #185 opened 3 years ago by MedadRufus
  0

• Limit dynamics info to the resolution of google maps

  #186 opened 3 years ago by MedadRufus
  0

• Is distance step a feature? or slope

  #188 opened 3 years ago by MedadRufus
  0

• Train a model to predict current on different surfaces, locations

  #142 opened 3 years ago by MedadRufus
  0

• Fix all the units on the plots

  #156 opened 3 years ago by MedadRufus
  0

• plot predicted vs actual, in 2 axis

  #174 opened 3 years ago by MedadRufus
  1

• Compare offline predictions with all the data I have

  #178 opened 3 years ago by MedadRufus
  0

• Find out what the real motor drag coefficient

  #147 opened 3 years ago by MedadRufus
  6

• Use LSTM and compare models

  #158 opened 3 years ago by MedadRufus
  1

• Just do test train split with no shuffle. Shuffle the dataframe before merging

  #179 opened 3 years ago by MedadRufus
  0

• Use the xgboost score, to compare modules

  #176 opened 3 years ago by MedadRufus
  0

• Train the model to predict energy usage over a stretch rather than power

  #144 opened 3 years ago by MedadRufus
  0

• Identify if the ML/Model over or underestimates Energy consumption along route

  #160 opened 3 years ago by MedadRufus
  1

• Train Autokeras image regressor with overlapping sections of data.

  #154 opened 3 years ago by MedadRufus
  0

• Run autokeras image regressor without SOC feature

  #153 opened 3 years ago by MedadRufus
  0

• Measure vertical velocity/Acceleration as a seperate feature.

  #161 opened 3 years ago by MedadRufus
  1

• Create data shape like this for Keras

  #150 opened 3 years ago by MedadRufus
  1

• Do masters project outline of paper

  #152 opened 3 years ago by MedadRufus
  0

• Add FIR speed as a featuer low passed, to capture history, inaddition to current speed.

  #169 opened 3 years ago by MedadRufus
  0

• Do a comparison between test data that has been ridden, vs a completely new route

  #165 opened 3 years ago by MedadRufus
  0

• Cannot extrapolate when using trees, vs linear/NN

  #172 opened 3 years ago by MedadRufus
  1

• Do a comparison in accuracy with and without the dynamics parameters

  #173 opened 3 years ago by MedadRufus
  0

• Test by each trip

  #166 opened 3 years ago by MedadRufus
  0

• Consider making distribution plots like these

  #167 opened 3 years ago by MedadRufus
  0

• Use standard scalar for the Deep ML models

  #168 opened 3 years ago by MedadRufus
  0

• The special iffy parameters are: vertical and horizontal speed, acceleration on surface

  #164 opened 3 years ago by MedadRufus
  0

• split the tests into 15 minute rides, and assess the energy consumption accuracy on all those rides

  #170 opened 3 years ago by MedadRufus
  0

• Try model based approach vs Machine Learning.

  #157 opened 3 years ago by MedadRufus
  0

• Consider using GPS speed as a feature.

  #159 opened 3 years ago by MedadRufus
  1

• Proove good correlation between GPS speed and internal speed

  #162 opened 3 years ago by MedadRufus
  1

• Make test train split return a random continueous set of rows for testing.

  #163 opened 3 years ago by MedadRufus
  0

• Instead of mergingasof to 10ms intervals, merge to the GPS frequency

  #155 opened 3 years ago by MedadRufus
  0

• Find out realistically, how high should the sample rate be to make predictions accurate.

  #183 opened 3 years ago by MedadRufus
  1

• Use a metric to determine if the test data has been seen in training

  #182 opened 3 years ago by MedadRufus
  0

• Create plots of distributions of all the data features

  #149 opened 3 years ago by MedadRufus
  0

• Start of trip SOC can be determined from simple OCV

  #143 opened 3 years ago by MedadRufus
  0

• Use Heading as a feature

  #145 opened 3 years ago by MedadRufus
  0

• Use GPS Speed as feature

  #146 opened 3 years ago by MedadRufus
  0

• Calculate SOC as a function of {Voltage, Current}

  #148 opened 3 years ago by MedadRufus
  0

• Predict SOC seperately

  #141 opened 3 years ago by MedadRufus
  0

• Plot SOC vs Power

  #140 opened 3 years ago by MedadRufus
  0