In the same directory level as MBTAriderSegmentation, run the following code. The extracted features would be saved to MBTAriderSegmentation/data/cached_features directory by default. The default paths can be changed in MBTAriderSegmentation/config.py
import time
from MBTAriderSegmentation.config import *
from MBTAriderSegmentation.features import FeatureExtractor
import time
t0 = time.time()
start_month = '1710' # specify start month
duration = 1 # specify length of duration
extractor = FeatureExtractor(start_month=start_month, duration=duration).extract_features()
print("feature extraction time: ", time.time() - t0)
In the same directory level as MBTAriderSegmentation, run the following code. The clustering results would be saved to MBTAriderSegmentation/data/cached_clusters directory by default. The default paths can be changed in MBTAriderSegmentation/config.py
import time
from MBTAriderSegmentation.config import *
from MBTAriderSegmentation.segmentation import Segmentation
# Hierarchical pipeline
import time
t0 = time.time()
start_month = '1710'
duration = 1
segmentation = Segmentation(start_month=start_month, duration=duration)
segmentation.get_rider_segmentation(hierarchical=True)
print("Hierarchical clustering time: ", time.time() - t0)
# Non-hierarchical pipeline
t0 = time.time()
start_month = '1710' # specify start month
duration = 1 # specify length of duration
segmentation = Segmentation(start_month=start_month, duration=duration)
segmentation.get_rider_segmentation(hierarchical=False)
print("Non-hierarchical clustering time: ", time.time() - t0)
In the same directory level as MBTAriderSegmentation, run the following code. The profiled cluster summary files would be saved to MBTAriderSegmentation/data/cached_profiles directory by default. The default paths can be changed in MBTAriderSegmentation/config.py
import time
from MBTAriderSegmentation.config import *
from MBTAriderSegmentation.profile import ClusterProfiler
# by_cluster = True
start_month = '1710' # specify start month
duration = 1 # specify length of duration
hier_flags = [True, False]
algo_types = ['kmeans', 'lda']
# by_cluster = True
for hier_flag in hier_flags:
for algo_type in algo_types:
t0 = time.time()
profiler = ClusterProfiler(start_month=start_month, duration=duration, hierarchical=hier_flag)
profiler.extract_profile(algorithm=algo_type, by_cluster=True)
print("[by_cluster = True] Profile time: ", time.time() - t0)
# by_cluster = False
t0 = time.time()
profiler = ClusterProfiler(start_month=start_month, duration=duration, hierarchical=True)
profiler.extract_profile(algorithm='kmeans', by_cluster=False)
print("[by_cluster = False] Profile time: ", time.time() - t0)
In the same directory level as MBTAriderSegmentation, run the following code. The visualizations for temporal patterns, cluster pattern-of-use and demographics plots would pop up and the map visualization for the geographical features would be saved as html files to MBTAriderSegmentation/data/cached_viz directory by default. The default paths can be changed in MBTAriderSegmentation/config.py
import time
from MBTAriderSegmentation.config import *
from MBTAriderSegmentation.visualization import Visualization
# load data
start_month = '1710' # specify start month
duration = 1 # specify length of duration
viz = Visualization(start_month=start_month, duration=duration)
viz.load_data(by_cluster=True, hierarchical=True, w_time=None, algorithm='lda')
# plot cluster_size, avg_num_trips
viz.plot_cluster_size()
viz.plot_avg_num_trips()
# PCA plot
viz.visualize_clusters_2d()
# plot temporal patterns
viz.plot_all_hourly_patterns()
# generate geographical patterns as html files
unique_clusters = list(viz.df['cluster'].unique())
map_urls = []
for cluster in unique_clusters:
map_urls.append(viz.plot_cluster_geo_pattern(cluster))
# plot demographics distribution
viz.plot_demographics(grp='race', stacked=True)
viz.plot_demographics(grp='edu', stacked=False)