Analyzing Demand Convergence

In this example we show how to use analyze demand convergence for across model iterations

Imports

from pathlib import Path

import numpy as np
from polaris.analyze.trip_metrics import TripMetrics
from tqdm import tqdm

Data Sources

Open the demand database for analysis

model_fldr = Path("/tmp/Bloomington")
supply_db = model_fldr / "Bloomington-Supply.sqlite"

demand_file = "Bloomington-Demand.sqlite"

iterations = 4
demand_files = {f"iter_{i}": model_fldr / f"Bloomington_iteration_{i}" / demand_file for i in range(1, iterations + 1)}

Data retrieval

data = {}

for k, file_name in tqdm(demand_files.items()):
    tm = TripMetrics(supply_file=supply_db, demand_file=file_name)
    data[k] = tm.data

  0%|          | 0/4 [00:00<?, ?it/s]
 25%|██▌       | 1/4 [00:00<00:01,  2.42it/s]
 50%|█████     | 2/4 [00:00<00:00,  2.20it/s]
 75%|███████▌  | 3/4 [00:01<00:00,  2.86it/s]
100%|██████████| 4/4 [00:01<00:00,  3.29it/s]
100%|██████████| 4/4 [00:01<00:00,  2.95it/s]

Plotting Trip Length distribution

import matplotlib.pyplot as plt

fig, ax = plt.subplots(figsize=(10, 6))

for k, df in data.items():
    dfa = df[df.travel_distance > 0]

    # For large models, using 200 to 400 bins yields much better curves
    y, x = np.histogram(dfa.travel_distance.values / 1000, bins=90)
    ax.plot(x[1:], y, label=k)

ax.legend(loc="upper right")
plt.title("Trip length distribution", fontsize=24)
plt.xlabel("km", fontsize=16)
plt.ylabel("Trips", fontsize=16)
plt.xlim(0, 25)
plt.tight_layout()
# fig.savefig("D:/my_project/TLFD.png")
plt.show()

Plotting travel departure distribution

import matplotlib.pyplot as plt

fig, ax = plt.subplots(figsize=(10, 6))
for k, df in data.items():
    y, x = np.histogram(df.tstart.values, bins=100)
    ax.plot(x[1:], y, label=k)

ax.legend(loc="upper left")
plt.title("Trip departure distribution", fontsize=24)
plt.xlabel("Seconds", fontsize=16)
plt.ylabel("Trips", fontsize=16)
plt.show()

Plotting correlation for origins and destinations to zones for successive iterations

def correlate_iterations(df1, df2):
    df1a = df1.rename(columns={"origins": "o_a", "destinations": "d_a"})
    df2a = df2.rename(columns={"origins": "o_b", "destinations": "d_b"})
    df = df1a.join(df2a).fillna(0)
    return np.corrcoef(df["o_b"].values, df["o_a"].values)[0, 1], np.corrcoef(df["d_b"].values, df["d_a"].values)[0, 1]


trip_vectors = {}
for i in range(iterations):
    df = data[f"iter_{i + 1}"]
    df1 = df.groupby(["origin"]).count()[["trip_id"]].rename(columns={"trip_id": "origins"})
    df2 = df.groupby(["destination"]).count()[["trip_id"]].rename(columns={"trip_id": "destinations"})
    trip_vectors[f"iter_{i + 1}"] = df1.join(df2).fillna(0)

origins = []
destinations = []
for i in range(iterations - 1):
    df1 = trip_vectors[f"iter_{i + 1}"]
    df2 = trip_vectors[f"iter_{i + 2}"]
    o, d = correlate_iterations(df1, df2)
    origins.append(o)
    destinations.append(d)

import matplotlib.pyplot as plt

fig, ax = plt.subplots(figsize=(10, 6))
plt.plot(origins)
plt.plot(destinations)
plt.ylim(0.9, 1.05)
plt.title("Correlation between successive iterations for both origins and destinations", fontsize=16)

Text(0.5, 1.0, 'Correlation between successive iterations for both origins and destinations')