Activity for Chickenmarkus
-
Chickenmarkus committed [r325]
Movebis moved to https://bitbucket.org/tudresden/preprocessing_data_science/
-
BearingResampleTrack added (calculation copied from Daniel).
-
Running docker containers as user including its iPython history.
-
Working resampling with overlap.
-
Interface prepared for the resampling with overlapping windows.
-
InterpolatedTrack added.
-
Coexistence of old pandas (before commit 313) and new pandas (with commit 313).
-
Docker: ipython3 splitted from python3-datascience.
-
Accerlation to LocFile added.
-
Normalization renamed to standardization (NormalizedTrack-->StandardizedTrack, etc.).
-
Write ipython-history into /tmp/ for docker.
-
Limited feature significance for 5s.
-
Updated to pandas>=0.24. Thus, the running version via docker on the vidserver.
-
Dockered iPython for laptop.
-
Executing the scripts in docker containers on the server.
-
make-script added.
-
03_features.py with the new resampled features.
-
DistanceResampledTrack for cumulative and haversine distance added.
-
Handling of the raw data from the CSV files more 'pandarized'.
-
StillstandResampledTrack added.
-
Variance added to MeanStdResampledTrack and range added to MinMaxResampledTrack.
-
PcTrack finished.
-
PcTrack in progress.
-
Feature variance as described for Otsu on Wikipedia.
-
Feature variance based on the samples of the groups.
-
LabelSegementedTrack fixed for robust (start,end)-times. Data types of QuantileResampledTrack fixed (unstack() --> one Series with a dtype for all columns).
-
Replace unset levels in MultiIndex of columns by an empty string ''.
-
The start of loading the features of all tracks added.
-
Several *Tracks are more robust especially in cases of resampling intervals of no data.
-
Distance and, hence, speed_calculated fixed.
-
Units changed to SI-units in the raw file data. ToDo: speed_device differs significantly from speed_calculated.
-
The global default and the individual normalization data frame in NormalizedTrack fixed.
-
NormalizedTrack updated to MultiIndex.
-
HistResampledTrack sketched. ToDo: How to get column-specific histogram bins?
-
Drop NaNs in QuantileResampledTrack. Delta_norm fixed in XyzFile.
-
Class diagram updated.
-
QuantileResampledTrack added.
-
MinMaxResampledTrack added.
-
Working FrequencyResampledTrack! *yeah*
-
02_trip.py updated for new StillstandTrack.
-
Class layout for resampling re-concepted for stackability.
-
Typos Segement-Segment.
-
LabelSegmentedTrack added.
-
ResampledTrack with reusing the resampling object by selecting the columns. Support of Categories.
-
Lable added. To do: Resampling a) Category and b) MultiIndex (https://stackoverflow.com/q/57873813/2053808).
-
Re-arranged concept of track segmentation and TrackSegments.
-
Concept changed for segmentation by label.
-
Additional layer 'AbstractData' above AbstractTrack added in order to add [Abstract]Features.
-
MovementTrack changed to non-segmented StillstandTrack.
-
SegementedTrack can be splitted into single Trips.
-
Iteration through the trips added to SegmentedTrack.
-
SegmentedTrack added.
-
Trip overview with new FrequencyTrack. To Do: TrackProcessor for trip segmentation by loc_speed_device_mean<=x and acc_norm_std<=y.
-
Potentially working FrequencyTrack with correct fftfreq and column wise FFT.
-
FrequencyTrack completed.
-
Work in progess for FrequencyTrack. To do: concatenate MultiIndexed DataFrames
-
Deprecated super class added. Violet UML diagram added.
-
Pandas.MultiIndex introduced. To do: FrequencyTrack.
-
FrequencyTrack sketched.
-
Sketch of LengthAutocorrelatedTrack with the additional length domain to the frequency domain added.
-
Possibility to limit the compute intensive calculation of AutocorrelatedTrack to specific columns.
-
XyzFile limited to the norm and its delta.
-
Autocorrelation/AutocorrelatedTrack added.
-
Structure changed from decorators to processors. TODO: ClassVariableForwarder.
-
Trip updated to the new track package. TODO: Segementation.
-
Nicely readably output for the PCA-matrix added.
-
PcTrackDecorator class added.
-
Working PCA analysis.
-
PCA debugging finished. Next step: find appropriate columns.
-
NormalizedTrackDecorator changed to a global normalization. TODO: Column handling due to NaN values in 01_pca.py.
-
01_pca.py modified for the new tracks with all sensor data. ToDo: Skipping of many tracks due to TypeError(ufunc 'isnan' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule 'safe')
-
ResampledTrack, TrackDecorator and NormalizedTrackDecorator added.
-
PropertyMeta replaced by __getitem__ so that child classes do not inherit the outdated properties of its parent.
-
To do: ResampledTrack.
-
GeneralTrack and Track with a common DataFrame added. Index of cls.data changed to pandas.DatetimeIndex.
-
Properties for each column in data moved to CsvFile class.
-
Startet to re-create the Track class for tracks with all sensor data.
-
track.py moved to a new package named track.legacy.
-
Using the entire dataset except the delta_time for evaluation.
-
Bugfix for incorrectly calculated distance and, hence, speed.
-
PCATrack and SegementedPCATrack moved to the track-file.
-
Trip segementation can be done with more than one component.
-
Movebis added.
-
Matrix and ordered statistics for quantile of 0.05 and 0.04.
-
Plot average OC of the consolidation simulation.
-
Save covariance matrix when calculating OC-matrix.
-
Save covariance matrix when calculating OC-matrix.
-
Extending the 3D-plot by a 2D-plot.
-
Wider range of OCmax.
-
OC-FFD with ordering by i) mean and ii) OC.
-
ND-CSP evaluation: using AVG(c.o.p.) instead of MAX(c.o.p.).
-
Saving evaluation results of the OC-CSP added.
-
FFD-OC evaluation with fits.
-
ND-FFD simulation plotting completed. Fixing incorrect overlap coefficients in Instance.
-
Running FFD-Simulation consolidation and corresponding evaluation.
-
Simulation completed and started to debug. However, Instance has overlap_coefficient\(\) when single-threaded but not when multiprocessed\?\!\?
-
Matrix rollbacked to correlation based.
-
Simulation of the ND-FFFD with respect to the overlap coefficient sketched.
-
Debugging message fixed.
-
Ordering of the task_usage_statistics based on the OC-matrix added.
1 >