AI Logic and Introduction to Machine Learning in Mapo, Seoul

1. Understanding Time Series Data

Time series data records events in chronological order. Date columns need proper parsing—not treatment as plain strings—so you can resample, trend, and compare periods.

2. Setting a Datetime Index

import pandas as pd

df = pd.read_csv("ecommerce_orders.csv", parse_dates=["Order Date"])
df = df.set_index("Order Date")

print("Dataset Shape:", df.shape)
print("Index Type:", type(df.index))
print(df.head(3))

With Order Date as the index, rows become chronological records of when orders arrived.

3. Monotonicity Check

is_ordered = df.index.is_monotonic_increasing
print("Datetime Index is Monotonically Increasing:", is_ordered)

If this is True, timestamps do not jump backward—useful before aggregations.

4. Monthly Resampling with 'ME'

Modern Pandas uses 'ME' (Month End) instead of the older 'M' alias:

monthly_revenue = df["Revenue"].resample("ME").sum()
print(monthly_revenue.head())

Illustrative totals: Jan 2022 ≈ 35,624; Feb 2022 ≈ 29,998 (indexed to each month’s last calendar day).

5. Month-over-Month Growth (.pct_change)

monthly_growth = monthly_revenue.pct_change()
print(monthly_growth.head())

First period is NaN (no prior month). Example: Feb ≈ −16% (shorter month); Mar ≈ +1.3%; Apr ≈ +1.5%. Positive values mean growth; negative values mean contraction.

6. Aggregation Sanity Check

raw_revenue_sum = df["Revenue"].sum()
resampled_revenue_sum = monthly_revenue.sum()
print("Match:", resampled_revenue_sum == raw_revenue_sum)

If both sums match, resampling did not drop or double-count rows.