Unlock E‑Commerce Insights: Python‑Driven Taobao Data Analysis Walkthrough

1. Project Background and Analysis

Project background: e‑commerce is essential; this project uses Taobao app data to analyze user behavior and discover patterns.

2. Data and Field Description

The dataset covers 2014‑11‑18 to 2014‑12‑18, mobile Taobao, 12,256,906 records, 6 columns: user_id , item_id , behavior_type (1‑click, 2‑collect, 3‑cart, 4‑pay), user_geohash , item_category , time .

3. Analysis Dimensions

Traffic metrics analysis

User behavior analysis

Funnel loss analysis

User value RFM analysis

2. Import Libraries

import numpy as np
import pandas as pd
import matplotlib as mpl
import matplotlib.pyplot as plt
import seaborn as sns
import warnings
sns.set(style="darkgrid", font_scale=1.5)
mpl.rcParams["font.family"] = "SimHei"
mpl.rcParams["axes.unicode_minus"] = False
warnings.filterwarnings('ignore')

3. Data Preview and Pre‑processing

df = pd.read_csv("taobao.csv", dtype=str)
df.shape
df.info()
df.sample(5)

Calculate missing rate, drop user_geohash , split time into date and hour , convert types, sort by time, and reset index.

# calculate missing rate
missing_rate = df.apply(lambda x: sum(x.isnull())/len(x), axis=0)
# drop geohash column
df.drop(["user_geohash"], axis=1, inplace=True)
# split time column
df["date"] = df.time.str[0:-3]
df["hour"] = df.time.str[-2:]
# convert types
df["date"] = pd.to_datetime(df["date"])
df["time"] = pd.to_datetime(df["time"])
df["hour"] = df["hour"].astype(int)
# sort and reset index
df.sort_values(by="time", ascending=True, inplace=True)
df.reset_index(drop=True, inplace=True)

4. Model Construction

Compute total page views (PV) and unique visitors (UV), daily and hourly metrics, behavior type counts, funnel analysis, and RFM segmentation.

# total PV and UV
total_pv = df["user_id"].count()
total_uv = df["user_id"].nunique()
# daily PV/UV
pv_daily = df.groupby("date")["user_id"].count()
uv_daily = df.groupby("date")["user_id"].apply(lambda x: x.nunique())
# hourly PV/UV
pv_hour = df.groupby("hour")["user_id"].count()
uv_hour = df.groupby("hour")["user_id"].apply(lambda x: x.nunique())
# behavior type counts
type_1 = df[df["behavior_type"] == "1"]["user_id"].count()
type_2 = df[df["behavior_type"] == "2"]["user_id"].count()
type_3 = df[df["behavior_type"] == "3"]["user_id"].count()
type_4 = df[df["behavior_type"] == "4"]["user_id"].count()

Funnel analysis merges click, collect, and cart into a single stage to show conversion rates.

# funnel data preparation
df_count = df.groupby("behavior_type").size().reset_index(name="人数")
type_dict = {"1":"点击","2":"收藏","3":"加入购物车","4":"支付"}
df_count["环节"] = df_count["behavior_type"].map(type_dict)
a, b, c, d = df_count["人数"]
# combine collect and cart
funnel = pd.DataFrame({"环节":["点击","收藏及加入购物车","支付"],"人数":[a, b + c, d]})
funnel["总体转化率"] = funnel["人数"] / funnel["人数"].iloc[0]
funnel["单一转化率"] = [1.0, (b + c) / a, d / (b + c)]

RFM analysis calculates recent purchase days, purchase frequency, assigns scores, and labels customers.

# simplified RFM calculation
recent_buy = df[df["behavior_type"] == "4"].groupby("user_id")["date"].max()
freq = df[df["behavior_type"] == "4"].groupby("user_id")["date"].count()
rfm = pd.concat([recent_buy, freq], axis=1).rename(columns={"date":"recent", "date":"freq"})
# scoring and labeling steps are omitted for brevity

All visualizations are generated with matplotlib , seaborn , or plotly (code omitted for brevity).