Hi @adamschroeder, Indeed, upon exploring the dataset, I see that for the Random Forest model, “Defensive Driving” is the variable that best separates approved from not approved cases. That’s why it appears near the root of the tree and acts as a good discriminator. However, this could be confirmed with a deeper analysis.
1 Like
@adamschroeder Thank you for your feedback! 
I initially wanted to put the filters on the left side and the cards on top, but I didn’t like it because the stack bar chart needed more space and I wanted everything to be visible. I practiced the layout with this, which took me quite a while to get the charts and cards to be in line at the bottom, and it ended up like this.
2 Likes
This week’s data set has 5 requirements for getting a New York City taxi medallion, which is a transferable permit that allows taxicabs to operate.
The requirements are:
DEFENSIVE_DRIVING: New York State certified 6-hour defensive driving course
DRIVER_EXAM: 70% score required on final exam
DRUG_TEST: must be taken at an approved facility. Nobody wants a driver under the influence.
MEDICAL_CLEARANCE: drivers should pass a minimal health screening
WAV_COURSE: Wheelchair Accessible Vehicle training
Polars is used to group the data by applicants who have completed 0, 1, 2, 3, 4, or 5 of these requirements. I wanted to see what is going on with applicants who completed 4 of the 5. Wondered which requirement is the last one for most. The dash bootstrap radio buttons provide the filter value for each displayed visualization.
When 0 is selected, all 5 bars are solid red, when 5 is selected all 5 bars are solid green. Here is the view of applicants who completed 4 of the 5 requirements. WAV_COURSE was completed by 464 of 466 (maybe the easiest or quickest of the 5?), DRIVER_EXAM was the most incomplete item of applicants with 4 of 5 requirements met.
I tried very hard to deploy this dashboard on pycafe, seems it is not set for directly using polars.
Here is the code:
import polars as pl
import plotly.express as px
from dash import Dash, dcc, html, Input, Output
import dash_bootstrap_components as dbc
#----- GLOBAL DATA STRUCTURES --------------------------------------------------
eval_cols = [
'DRUG_TEST','WAV_COURSE','DEFENSIVE_DRIVING', 'DRIVER_EXAM',
'MEDICAL_CLEARANCE'
]
color_dict = {
'COMPLETE' : 'mediumspringgreen',
'INCOMPLETE' : 'lightcoral'
}
style_space = {
'border': 'none',
'height': '4px',
'background': 'linear-gradient(to right, #007bff, #ff7b00)',
'margin': '10px,',
'fontsize': 32
}
source_data = (
'https://raw.githubusercontent.com/plotly/Figure-Friday/refs/heads/' +
'main/2025/week-19/TLC_New_Driver_Application.csv'
)
#----- CALLBACK FUNCTION ---- --------------------------------------------------
def get_bar_chart(df, completed_val):
data_len = dict_complete_len.get(completed_val)
df_bar = (
df
.select(pl.col('REQUIREMENT',completed_val))
.rename({completed_val:'COMPLETE'})
.with_columns(
INCOMPLETE = data_len - pl.col('COMPLETE')
)
)
bar_chart = px.bar(
df_bar,
y='REQUIREMENT',
x=['COMPLETE', 'INCOMPLETE'],
# height=600, width=900,
template='simple_white',
title=(
f'COMPLETED {completed_val} of {requirement_num} REQUIREMENTS<br>' +
f'<sup>Status of {data_len} Applicants</sup>'
),
color_discrete_map=color_dict,
)
bar_chart.update_layout(
yaxis=dict(autorange="reversed"),
font_size=20,
legend_title_text='Status'
)
bar_chart.update_xaxes(title_text='APPLICANTS')
bar_chart.update_yaxes(title_text='')
return(bar_chart)
#----- READ & CLEAN DATASET ----------------------------------------------------
df = (
pl
.scan_csv(source_data)
.select(
APP_DATE = pl.col('App Date').str.to_date(format='%m/%d/%Y'),
STATUS = pl.col('Status')
.str.replace('Approved - License Issued', 'Approved')
.str.replace('Pending Fitness Interview', 'Fitness_Interview'),
DRUG_TEST = pl.when(pl.col('Drug Test') == 'Complete')
.then(pl.lit(1))
.otherwise(pl.lit(0)),
WAV_COURSE = pl.when(pl.col('WAV Course') == 'Complete')
.then(pl.lit(1))
.otherwise(pl.lit(0)),
DEFENSIVE_DRIVING = pl.when(pl.col('Defensive Driving') == 'Complete')
.then(pl.lit(1))
.otherwise(pl.lit(0)),
DRIVER_EXAM = pl.when(pl.col('Driver Exam') == 'Complete')
.then(pl.lit(1))
.otherwise(pl.lit(0)),
MEDICAL_CLEARANCE = pl.when(pl.col('Medical Clearance Form') == 'Complete')
.then(pl.lit(1))
.otherwise(pl.lit(0)),
)
.filter(pl.col('APP_DATE') >= pl.datetime(2024,1,1)) # exclude pre-2024
# count occurances of 'Complete in 5 listed columns
.with_columns(COMPLETES = pl.sum_horizontal(eval_cols))
.select(
DRUG_TEST=pl.col('DRUG_TEST')
.sum()
.over('COMPLETES'),
WAV_COURSE=pl.col('WAV_COURSE')
.sum()
.over('COMPLETES'),
DEFENSIVE_DRIVING=pl.col('DEFENSIVE_DRIVING')
.sum()
.over('COMPLETES'),
DRIVER_EXAM=pl.col('DRIVER_EXAM')
.sum()
.over('COMPLETES'),
MEDICAL_CLEARANCE=pl.col('MEDICAL_CLEARANCE')
.sum()
.over('COMPLETES'),
COMPLETES = pl.sum_horizontal(eval_cols)
)
.with_columns(
LEN=pl.col('DRUG_TEST')
.count()
.over('COMPLETES'),
)
.with_columns(pl.col('COMPLETES').cast(pl.String))
.unique('COMPLETES')
.sort('COMPLETES')
.collect()
)
dict_complete_len = dict(zip(df['COMPLETES'], df['LEN']))
completes_list = list(dict_complete_len.keys())
df_transpose = (
df
.transpose(
include_header=True,
column_names='COMPLETES',
header_name='REQUIREMENT',
)
.filter(pl.col('REQUIREMENT') != 'LEN')
.sort('REQUIREMENT')
)
requirement_num = df_transpose.height
#----- DASH APPLICATION STRUCTURE-----------------------------------------------
app = Dash(external_stylesheets=[dbc.themes.LUX])
app.layout = dbc.Container([
html.Hr(),
html.H2(
'NYC Taxi Application Status',
style={'text-align': 'center', 'font-size': '32px'}
),
html.Hr(style=style_space),
html.Div([
html.P(
'Radio buttons select # of completed requirements',
style={
'text-align': 'center',
'margin-left': '80px',
'margin-right': '80px',
'font-size': '32px',
'color': 'gray',
}
),
html.Hr(style=style_space)
]),
dbc.Col(
[
dbc.RadioItems(
id='radio_sel',
options=completes_list,
value = '4',
inline=True,
labelStyle={
'margin-right': '30px',
'font-size': '20px',
}
)
]
),
html.Div(id='dd-output-container'),
dbc.Row([
dbc.Col(dcc.Graph(id='bar_chart')),
]),
])
@app.callback(
Output('bar_chart', 'figure'),
Input('radio_sel', 'value'),
)
def update_dashboard(completed_val):
return (
get_bar_chart(df_transpose, completed_val)
)
if __name__ == '__main__':
app.run_server(debug=True)
3 Likes