def get_segment_traces(group, segment):
color = display.COLORS[group.index % len(display.COLORS)]
return plotting.make_segment_traces(
pes_tracks,
segment,
color=color,
name='Gauge #{}'.format(group.index + 1),
show_legend=(group.segments.index(segment) < 1),
legend_group='gauge-group-{}'.format(group.index))
def compute_unweighted(
tracks: pd.DataFrame,
tolerance_fraction: float
) -> dict:
""" """
pes_tracks = tracks[tracks['pes'] == 1].copy() # type: pd.DataFrame
tolerance = 0.5 * tolerance_fraction * np.median(pes_tracks['width'].values)
pes_tracks['tolerance'] = tolerance
scatter_trace = plotting.create_scatter(
pes_tracks['curvePosition'],
pes_tracks['simpleGauge'],
color='rgba(200, 200, 200, 1)',
name='Gauges'
)
segments = forward_windowing.compute(
pes_tracks['simpleGauge'].tolist(),
pes_tracks['tolerance'].tolist(),
weighted=False
)
segment_traces = [
trace
for s in segments
for trace in plotting.make_segment_traces(pes_tracks, s)
]
return dict(
segments=segments,
data=[scatter_trace] + segment_traces,
layout=go.Layout(
xaxis={'title': 'Trackway Position (m)'},
yaxis={'title': 'Gauge (m)'}
)
)
layout = go.Layout(
title='{} Pes Gauge Values Along Trackway'.format(trackway_name),
xaxis={'title': 'Trackway Position (m)'},
yaxis={'title': 'Gauge (m)'})
scatter_trace = plotting.create_scatter(pes_tracks['curvePosition'],
pes_tracks['simpleGauge'],
pes_tracks['simpleGaugeUnc'],
color='rgba(200, 200, 200, 1)',
name='Gauges')
segments = forward_windowing.compute(pes_tracks['simpleGauge'].tolist(),
pes_tracks['simpleGaugeUnc'].tolist())
segment_traces = [
trace for s in segments
for trace in plotting.make_segment_traces(pes_tracks, s)
]
cd.display.markdown("""
## Segmentation with Uncertainties
The first, and most crucial, change to make to the analysis is to introduce
uncertainties into the clustering process. Their introduction allows us to
segment using statistical significance instead of an arbitrary assigned
tolerance.
""")
cd.display.plotly(data=[scatter_trace] + segment_traces, layout=layout)
cd.display.markdown("""
The result of uncertainty-based clustering differs substantially from
tracks['simpleGauge'],
tracks['simpleGaugeUnc'],
color='rgba(200, 200, 200, 1)',
name='Gauges'
)
forward_segments = forward_windowing.compute(
tracks['simpleGauge'].tolist(),
tracks['simpleGaugeUnc'].tolist()
)
cd.display.plotly(
data=[scatter_trace] + [
trace
for s in forward_segments
for trace in plotting.make_segment_traces(tracks, s)
],
layout=layout
)
layout = go.Layout(
title='Precision-Window Gauge Clustering',
xaxis={'title': 'Trackway Position (m)'},
yaxis={'title': 'Gauge (m)'}
)
scatter_trace = plotting.create_scatter(
tracks['curvePosition'],
tracks['simpleGauge'],
tracks['simpleGaugeUnc'],
color='rgba(200, 200, 200, 1)',