Compute Cumulative Sums and Build Stacked Charts with Pandas

Data Requirement

Analyze the proportion of usage durations for each product and draw a stacked chart ordered by year.

Data Preparation

Generate random data with the following Python code:

random_seed = 2022
length = 10000

phone_list = ['IPhone 13', '小米 13', '小米 12', 'IPhone 12', 'iQOO Neo6 SE', '华为P50E', 'OPPO Find X5 Pro', '一加10']
used_year_list = ['1年', '2年', '3年', '4年', '5年', '5年以上']
lost_money_list = [np.linspace(100, 500, 5, dtype=int),
                 np.linspace(400, 1000, 5, dtype=int),
                 np.linspace(600, 1500, 5, dtype=int),
                 np.linspace(800, 2000, 5, dtype=int),
                 np.linspace(1000, 3000, 5, dtype=int),
                 np.linspace(1500, 4000, 5, dtype=int)]

np.random.seed(random_seed)
phones = np.random.choice(phone_list, length)
used_year = np.random.choice(used_year_list, length, p=[0.34, 0.2, 0.2, 0.2, 0.05, 0.01])

def choose_money(choice_list, p=[0.3, 0.3, 0.2, 0.1, 0.1]):
    return np.random.choice(choice_list, 1, p=p)[0]

product_data = pd.DataFrame({'产品': phones, '使用时长': used_year})
product_data.drop(index=np.random.permutation(length)[:int(length * 0.1)], inplace=True)
product_data['贬值金额'] = product_data['使用时长'].map(lambda x: choose_money(lost_money_list[used_year_list.index(x)]))

Or load the pre‑generated gzip file:

# If the file is not in the current directory, use an absolute path
product_data = pd.read_csv('random_product_lost_money_data.gz', encoding='utf-8')

Approach

Group by product, count usage years with value_counts, then normalize and reshape.

product_data.groupby(['产品'])['使用时长'].value_counts()

Convert the multi‑index Series to a DataFrame and fill values:

result_df = pd.DataFrame(0, columns=product_data['使用时长'].unique(), index=product_data['产品'].unique())
used_year_df = product_data.groupby(['产品'])['使用时长'].value_counts()
for (i, j), value in used_year_df.items():
    result_df.loc[i, j] = value

Compute row‑wise percentages using NumPy broadcasting:

result_df.loc[:, result_df.columns] = result_df.values / result_df.values.sum(axis=1).reshape(len(result_df), 1)

Calculate cumulative sums for stacking:

result_df = result_df.cumsum(axis=1)
# or using expanding
result_df = result_df.expanding(axis=1).sum()

Alternative methods using groupby → unstack or pd.crosstab with normalize='index' are also shown:

product_data.groupby(['产品'])['使用时长'].value_counts(normalize=True).unstack().cumsum(axis=1)
pd.crosstab(product_data['产品'], product_data['使用时长'], product_data['使用时长'], aggfunc='count', normalize='index').cumsum(axis=1)

Plotting

Draw a horizontal stacked bar chart with Matplotlib, ordering usage years from longest to shortest:

from matplotlib import pyplot as plt
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.rcParams['axes.unicode_minus'] = False
plt.figure(figsize=(14, 6))
colors = ['red', 'green', 'orange', 'blue', 'pink']
for num, col in enumerate(['5年', '4年', '3年', '2年', '1年']):
    plt.barh(product_data.index, product_data[col].values * 100, color=colors[num], label=col)
plt.xlim(0, 110)
plt.xlabel('占比情况')
plt.legend()
plt.show()

The chart displays percentage shares of each usage duration per product.

Conclusion

By using cumulative sums and built‑in pandas functions, the stacked plot can be generated without explicit loops, simplifying the code while clearly showing the distribution of usage durations across products.