def prepare_dataset(garage_id):
#path = r'E:\Files\leeuwarden\leeuwarden_data'
parking_locations = pd.read_csv(base_file_location(
r'parkeerdata\leeuwarden_garage_parking_garage_gps.csv'),
sep=";",
decimal=",")
name = r"data\out-visitors_per_hour-garage" + str(
garage_id) + "_predict.csv"
# Load data
visitors_per_hour = pd.read_csv(base_file_location(name), sep=";")
matched_row = parking_locations.loc[parking_locations.garage_id ==
garage_id]
capacity = int(matched_row.capacity_value)
visitors_per_hour['perc'] = 100 * visitors_per_hour['cum_sum'] / capacity
X = visitors_per_hour.copy()
# drop the target_variable
#X = X.drop(['occupation', 'cum_sum'], axis=1)
X = X.drop(['cum_sum'], axis=1)
# drop id from feature table (overfitting)
X = X.drop(['index'], axis=1)
# drop visitors (explainable variable with respect to the binned-target variable)
X = X.drop('perc', axis=1)
# drop datetime (overfitting)
X = X.drop(['date'], axis=1)
# drop direct influencers
X = X.drop(['count_in', 'count_out', 'delta', 'month'], axis=1)
# drop weather data dates
X = X.drop(['YYYYMMDD', 'HH'], axis=1)
return X
def save_one_fig(df_predictions, df_garage_locations, prediction_date,
prediction_hour):
plt.clf()
plt.figure(figsize=(15, 15))
update_drawing(prediction_hour, 24, df_predictions, df_garage_locations,
prediction_date)
plt.savefig(base_file_location(r"data\parking_lwd.png"))
def process_file():
visit_info = Q.PriorityQueue()
cnt = 0
with open(base_file_location('locatus\locatusdata_bewerkt_filtered.csv'), 'r') as infile:
# skip header
next(infile)
for line in infile:
visitor, sensor, visit_timestamp = parse_line(line)
visit_info.put((visitor, visit_timestamp, sensor))
cnt += 1
if cnt % 500000 == 0:
print("cnt: {0}".format(cnt))
return visit_info
def main():
# Create the animation for several days.
fig, ax = plt.subplots(figsize=(20, 15))
# We need to have ffmpeg installed
plt.rcParams['animation.ffmpeg_path'] = ffmpeg_location()
AnimationWriter = matplotlib.animation.writers['ffmpeg']
writer = AnimationWriter(fps=15,
metadata=dict(artist='pca 2018'),
bitrate=1800)
network_ani = matplotlib.animation.FuncAnimation(fig,
update_drawing,
frames=120,
interval=500,
repeat=False)
print("saving")
network_ani.save(base_file_location(r'data\im.mp4'), writer=writer)
def save_one_fig(days):
plt.clf()
plt.figure(figsize=(20, 15))
update_drawing(days)
plt.savefig(base_file_location(r"data\visit_lwd.png"))
# Author: pca
# Date: 2018-06-26
# Version: 1.0
#
# Prerequisite: ffmpeg.exe is available on machine
import networkx as nx
import pandas as pd
from leeuwarden_kennis.config_data import base_file_location
from leeuwarden_kennis.config_data import ffmpeg_location
import collections
import matplotlib.pyplot as plt
import matplotlib.animation
# this is the file created in network_lwd_store_routes
df_visits = pd.read_csv(base_file_location(r'data\network_walks.csv'), sep=';')
# add date only column
df_visits['visit_date'] = pd.to_datetime(
pd.to_datetime(df_visits['time_from'], format='%Y-%m-%d').dt.date)
all_dates = sorted(df_visits['visit_date'].unique())
# get the sensor locations.
# this file can be found here: https://github.com/KennisnetwerkDataScience/Wifi-punten-in-Leeuwarden/tree/master/data
df_sensor_locations = pd.read_csv(base_file_location(r'data\opstelpunten.csv'),
sep=';',
decimal=',')
def calc_width(max_width, total_routes, weight):
network_walks = []
last_person = -1
last_location = -1
last_visit_time = datetime.datetime.now()
while not visits_pq.empty():
visit_record = visits_pq.get()
current_person = visit_record[0]
current_location = visit_record[2]
current_visit_time = visit_record[1]
# change of location
if current_location != last_location and current_person == last_person and in_time_hrs(last_visit_time, current_visit_time, 12):
network_walks.append((current_person, last_location, current_location, last_visit_time, current_visit_time))
last_person = current_person
last_location = current_location
last_visit_time = current_visit_time
return network_walks
visits_pq = process_file()
network = make_network(visits_pq)
# make a pandas df from this
network_df = pd.DataFrame(network, columns=['visitor', 'sensor_from', 'sensor_to', 'time_from', 'time_to'])
network_df.to_csv(base_file_location(r'data\network_walks.csv'), index=None, sep=';')
from sklearn.ensemble import RandomForestClassifier
import pickle
import networkx as nx
from leeuwarden_kennis.config_data import base_file_location
from leeuwarden_kennis.config_data import ffmpeg_location
import collections
import matplotlib.pyplot as plt
import matplotlib.animation
parking_ids = [36, 37, 38, 39, 40]
classifiers = {
garage_id: pickle.load(
open(
base_file_location(r"data\classifier_garage_{0}.sav").format(
garage_id), "rb"))
for garage_id in parking_ids
}
# prepare dataset
# fields:
# index;date;hour;count_in;count_out;delta;cum_sum;week;month;day;dayofyear;weekday;year;YYYYMMDD;HH;R
def create_dataset(prediction_date: datetime.datetime, rain: float):
pd_prediction_date = pd.to_datetime(prediction_date)
ds_dict = dict()
for idx in range(0, 24):
date = pd_prediction_date.strftime('%Y-%m-%d')
hr = idx