数据分析 - Pandas pylot的简单使用

# Import the necessary libraries

import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
sns.set(color_codes=True)

# Load the car.csv dataset
car = pd.read_csv('D:/KevinZhang/Study/data_analytics/USA_cars_datasets.csv')
car.head()
car = car.drop(columns=car.columns[0], axis='columns')
car.head()

car.dtypes

price             int64
brand            object
model            object
year              int64
title_status     object
mileage         float64
color            object
vin              object
lot               int64
state            object
country          object
condition        object
dtype: object

car.columns

Index(['price', 'brand', 'model', 'year', 'title_status', 'mileage', 'color',       'vin', 'lot', 'state', 'country', 'condition'],
      dtype='object')

# Analytical summary of the dataset
car.describe(include='all')

car.hist(figsize=(20,30))

array([[<AxesSubplot:title={'center':'price'}>,
        <AxesSubplot:title={'center':'year'}>],
       [<AxesSubplot:title={'center':'mileage'}>,
        <AxesSubplot:title={'center':'lot'}>]], dtype=object)

# Relationship between categorical and continuous variable
sns.boxplot(x='country', y='price', data=car)

<AxesSubplot:xlabel='country', ylabel='price'>

sns.pairplot(car)

<seaborn.axisgrid.PairGrid at 0x1a27976d520>

# Drop irrelevant columns
car = car.drop(['lot', 'condition'], axis=1)
car.head(5)

# Renaming the column names
car = car.rename(columns={'mileage': 'meter', 'title_status': 'type'})
car.head()

# Total number of rows and columns
car.shape

(2499, 10)

# Rows containing duplicate data
duplicate_rows_car = car[car.duplicated()]
print('Number of duplicate rows', duplicate_rows_car.shape)
duplicate_rows_car

Number of duplicate rows (4, 10)

# Count the rows before removing the data
car.count()

price      2499
brand      2499
model      2499
year       2499
type       2499
meter      2499
color      2499
vin        2499
state      2499
country    2499
dtype: int64

# Drop the duplicates
car = car.drop_duplicates()
car.count()

price      2495
brand      2495
model      2495
year       2495
type       2495
meter      2495
color      2495
vin        2495
state      2495
country    2495
dtype: int64

# Find the null values
print(car.isnull().sum())

price           0
brand           0
model           0
year            0
title_status    6
mileage         0
color           0
vin             0
lot             0
state           0
country         0
condition       4
dtype: int64

# Drop the missing values
car = car.dropna()
car.count()

price           2489
brand           2489
model           2489
year            2489
title_status    2489
mileage         2489
color           2489
vin             2489
lot             2489
state           2489
country         2489
condition       2489
dtype: int64

print(car.isnull().sum())

price           0
brand           0
model           0
year            0
title_status    0
mileage         0
color           0
vin             0
lot             0
state           0
country         0
condition       0
dtype: int64

# Finding the outliers 找出异常值
sns.boxplot(x=car['price'])

<AxesSubplot:xlabel='price'>

sns.boxplot(x=car['year'])

<AxesSubplot:xlabel='year'>

# Plotting a Histogram for number of cars per brand
car.brand.value_counts().nlargest(10).plot(kind='bar', figsize=(10,5))
plt.title('Number of cars by brand')
plt.xlabel('Brand')
plt.ylabel('Number of cars')

Text(0, 0.5, 'Number of cars')

# Find the relations between the variables
plt.figure(figsize=(20,10))
c = car.corr()
sns.heatmap(c, cmap='BrBG', annot=True)
c