Data Exploration and Preparation

Step 1: Load the Dataset

Download and load the California Fire Incidents dataset into Orange:

1. Download the dataset: California_Fire_Incidents.csv
2. In Orange, add a File widget to the canvas
3. Double-click the widget and select the downloaded CSV file
4. Connect a Data Table widget to view the data

Step 2: Explore the Data

Add visualization widgets to understand the data:

1. Connect a Distributions widget to see feature distributions
2. Connect a Scatter Plot widget to explore relationships
3. If you have geo add-on installed, use Geo Map to visualize fire locations

Questions to answer:

• What is the distribution of MajorIncident (True vs False)?
• Which counties have the most fires?
• Is there a correlation between AcresBurned and PersonnelInvolved?

Step 3: Handle Missing Values

Use the Impute widget to handle missing values:

1. Connect an Impute widget after your File widget
2. Choose imputation strategy: Average/Most frequent for numeric/categorical
3. Compare the data before and after imputation

Alternatively, use the Python Script widget for custom imputation:

from Orange.data.pandas_compat import table_from_frame, table_to_frame

df = table_to_frame(in_data)
# Fill numeric with median
df['AcresBurned'] = df['AcresBurned'].fillna(df['AcresBurned'].median())
# Fill categorical with 'Unknown'
df['Counties'] = df['Counties'].fillna('Unknown')
out_data = table_from_frame(df)

Step 4: Remove Duplicates

Use a Python Script widget to remove duplicate rows:

from Orange.data.pandas_compat import table_from_frame, table_to_frame

df = table_to_frame(in_data)
original_count = len(df)
df = df.drop_duplicates()
new_count = len(df)
print(f"Removed {original_count - new_count} duplicates")
out_data = table_from_frame(df)

Step 5: Handle Imbalanced Classes

The MajorIncident target is imbalanced (~66% False, ~34% True). Use undersampling:

import pandas as pd
from Orange.data.pandas_compat import table_from_frame, table_to_frame

df = table_to_frame(in_data)
# Separate by class
major = df[df['MajorIncident'] == True]
non_major = df[df['MajorIncident'] == False]
# Undersample majority class
non_major_sampled = non_major.sample(n=len(major), random_state=42)
# Combine
balanced_df = pd.concat([major, non_major_sampled])
print(f"Balanced dataset: {len(balanced_df)} rows")
out_data = table_from_frame(balanced_df)

Step 6: Feature Engineering - Date Features

Extract useful features from the Started date column:

import pandas as pd
from Orange.data.pandas_compat import table_from_frame, table_to_frame

df = table_to_frame(in_data)
# Convert to datetime
df['Started'] = pd.to_datetime(df['Started'], errors='coerce')
# Extract features
df['Month'] = df['Started'].dt.month
df['DayOfWeek'] = df['Started'].dt.dayofweek
df['Season'] = df['Month'].map({
    12:1, 1:1, 2:1,   # Winter
    3:2, 4:2, 5:2,    # Spring
    6:3, 7:3, 8:3,    # Summer
    9:4, 10:4, 11:4   # Fall
})
# Drop original date (prevents data leakage)
df = df.drop(columns=['Started', 'Extinguished'], errors='ignore')
out_data = table_from_frame(df)

Step 7: Build a Classification Model

Now build and evaluate a Random Forest model:

1. Add a Select Columns widget to set MajorIncident as target
2. Add a Random Forest widget (or other classifier)
3. Add a Test and Score widget to evaluate
4. Connect: Data → Select Columns → Test and Score ← Random Forest