Session Data Analysis.

Let's start with importing the dataset into the notebook.

Python
import pandas as pd

# Load dataset
df = pd.read_csv('sessions.csv')

Part 1: Calculate avg. session time

Python
# Filter sessions longer than 180 seconds
long_sessions = df[df['session_time'] > 180]

# Calculate average session time
average_session_time = long_sessions['session_time'].mean()
print("Average session time (longer than 3 minutes):", average_session_time)

Part 2: Visualize data

First, we must prepare the data for plotting.

Python
import matplotlib.pyplot as plt

# Create bins for 5-minute intervals (300 seconds)
bins = pd.interval_range(start=0, end=df['session_time'].max() + 300, freq=300)
df['bin'] = pd.cut(df['session_time'], bins=bins)

# Count sessions in each bin
hist_data = df['bin'].value_counts().sort_index()

# Prepare data for plotting
hist_df = pd.DataFrame({
    'bin': hist_data.index.astype(str),
    'count': hist_data.values
})

hist_df

Using this dataframe, we can then plot the histogram.

Python
# Plot histogram
plt.figure(figsize=(10, 6))
plt.bar(hist_df['bin'], hist_df['count'])
plt.xlabel('Session Length (seconds)')
plt.ylabel('Number of Sessions')
plt.title('Histogram of Session Lengths')
plt.xticks(rotation=45)
plt.show()

The histogram should look like this:

Part 3: Find similar countries

Next, we can do some pandas magic to aggregate the count of sessions for each country. Then, we find all the possible combinations of countries

Python
# Count sessions per country
session_counts = df['country'].value_counts().reset_index()
session_counts.columns = ['country', 'count']

# Find possible combinations
session_counts['key'] = 1
merged = session_counts.merge(session_counts, on='key', suffixes=('_a', '_b')).drop('key', axis=1)

Finally, we filter out those pairs that are not within 10% of each other.

Python
# Compute the similarity condition
condition = abs(merged['count_a'] - merged['count_b']) / merged[['count_a', 'count_b']].min(axis=1) <= 0.1

# Filter out self-pairs and non-similar pairs
similar_countries_df = merged[condition & (merged['country_a'] != merged['country_b'])][['country_a', 'country_b']]

# Drop duplicate pairs (A, B) and (B, A)
similar_countries_df = similar_countries_df[similar_countries_df.apply(lambda x: frozenset([x['country_a'], x['country_b']]), axis=1).duplicated(keep='first') == False]

similar_countries_df