Car Sales Analysis
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
sales = pd.read_csv("CarsData.csv")
test = pd.read_csv("test.csv",delimiter=";")
sales.head(20)
Makeobject
Toyota10%
GMC10%
16 others80%
Modelobject
Vitara5%
S20005%
18 others90%
0
Suzuki
Vitara
1
Honda
S2000
2
BMW
Z4
3
Toyota
Tacoma
4
Ford
Festiva
5
Buick
Skylark
6
Infiniti
QX
7
Ram
C/V
8
GMC
Safari
9
Nissan
Altima
test.head(20)
Makeobject
Chevrolet15%
Dodge15%
12 others70%
Modelobject
G5%
Suburban 25005%
18 others90%
0
Infiniti
G
1
Chevrolet
Suburban 2500
2
Mitsubishi
Precis
3
Mercedes-Benz
E-Class
4
Plymouth
Breeze
5
Chevrolet
Avalanche
6
Dodge
Grand Caravan
7
Toyota
Celica
8
Spyker
C8 Double 12 S
9
GMC
Yukon XL 1500
sales.columns
sales.shape
sales.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 4998 entries, 0 to 4997
Data columns (total 20 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Make 4998 non-null object
1 Model 4998 non-null object
2 Nickname 4998 non-null object
3 Car Gender 4998 non-null object
4 Buyer Gender 4998 non-null object
5 Buyer Age 4998 non-null int64
6 Buzzword 4998 non-null object
7 Country 4998 non-null object
8 City 4998 non-null object
9 Dealer Latitude 4998 non-null float64
10 Dealer Longitude 4998 non-null float64
11 Color 4998 non-null object
12 New Car 4998 non-null bool
13 Purchase Date 4998 non-null object
14 Sale Price 4998 non-null float64
15 Discount 4998 non-null float64
16 Resell Price 4998 non-null float64
17 5-yr Depreciation 4998 non-null float64
18 Top Speed 4998 non-null float64
19 0-60 Time 4998 non-null float64
dtypes: bool(1), float64(8), int64(1), object(10)
memory usage: 746.9+ KB
# Datos sobre las variables númericas
sales.describe()
Buyer Agefloat64
Dealer Latitudefloat64
count
4998
4998
mean
47.83313325
24.80011835
std
16.03538733
24.61560776
min
20
-46.5996116
25%
34
7.3938026
50%
48
30.9104435
75%
62
44.2729581
max
75
69.63186
# Datos sobre las variables categóricas
sales.describe(include = ["O"])
Makeobject
Modelobject
count
4998
4998
unique
69
876
top
Ford
Corvette
freq
428
31
sales.groupby(["Buyer Gender"]).count()['Model']
sales.groupby(["Buyer Gender","New Car"]).count()['Model']
grouped_sex = sales.groupby(['Buyer Gender', 'New Car']).count()['Model']
print(grouped_sex)
(grouped_sex.unstack(level=0).plot.bar())
plt.show()
Buyer Gender New Car
Female False 1252
True 1266
Male False 1267
True 1213
Name: Model, dtype: int64
grouped_sex.unstack(level=1)
falseint64
trueint64
Female
Male
print(sales.groupby(['Buyer Age', 'New Car'])
.count()["Model"]
.unstack(level=1)
.plot.bar(figsize=(20,5)))
AxesSubplot(0.125,0.125;0.775x0.755)
sales[[
"Make",
"Model",
"Buyer Gender",
"Buyer Age",
"New Car",
"Sale Price",
"5-yr Depreciation",
"Top Speed",
"0-60 Time"
]]
Makeobject
Ford8.6%
Chevrolet7.6%
67 others83.9%
Modelobject
Corvette0.6%
Camaro0.6%
874 others98.8%
0
Suzuki
Vitara
1
Honda
S2000
2
BMW
Z4
3
Toyota
Tacoma
4
Ford
Festiva
5
Buick
Skylark
6
Infiniti
QX
7
Ram
C/V
8
GMC
Safari
9
Nissan
Altima
sales[[
"Make",
"Model",
"Buyer Gender",
"Buyer Age",
"New Car",
"Sale Price",
"5-yr Depreciation",
"Top Speed",
"0-60 Time"
]].info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 4998 entries, 0 to 4997
Data columns (total 9 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Make 4998 non-null object
1 Model 4998 non-null object
2 Buyer Gender 4998 non-null object
3 Buyer Age 4998 non-null int64
4 New Car 4998 non-null bool
5 Sale Price 4998 non-null float64
6 5-yr Depreciation 4998 non-null float64
7 Top Speed 4998 non-null float64
8 0-60 Time 4998 non-null float64
dtypes: bool(1), float64(4), int64(1), object(3)
memory usage: 317.4+ KB
#metodo map del label encoding
sales['Buyer Gender'] = sales["Buyer Gender"].map({"Female": 1, "Male": 0}).astype(int)
sales['New Car'] = sales["New Car"].map({True: 1, False: 0}).astype(int)
sales[['Model', 'Buyer Gender', "New Car"]].head()
Modelobject
Buyer Genderint64
0
Vitara
1
1
S2000
0
2
Z4
1
3
Tacoma
0
4
Festiva
0
#sales.groupby(['Buyer Gender'])
sales.groupby(['Buyer Gender']).sum()["Sale Price"]
# Variables dependientes
Y_train = sales["Buyer Gender"]
# Preprocesamiento de variables independientes
X_train = sales[[
"Buyer Age",
"Top Speed",
"5-yr Depreciation",
"Sale Price"
]]
print(Y_train, X_train)
0 1
1 0
2 1
3 0
4 0
..
4993 0
4994 1
4995 1
4996 0
4997 0
Name: Buyer Gender, Length: 4998, dtype: int64 Buyer Age Top Speed 5-yr Depreciation Sale Price
0 51 200.9 0.13 54806.14
1 30 158.5 0.02 51826.30
2 54 149.5 0.24 82929.14
3 68 153.3 0.20 56928.66
4 70 122.0 0.18 77201.26
... ... ... ... ...
4993 51 191.7 0.11 67254.72
4994 38 243.1 0.24 68142.45
4995 23 224.8 0.12 57902.43
4996 62 201.6 0.24 57009.68
4997 49 129.9 0.07 71653.19
[4998 rows x 4 columns]
from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_train.shape
# Entrenar el modelo de árboles de descisión
from sklearn.tree import DecisionTreeClassifier
decision_tree = DecisionTreeClassifier()
decision_tree.fit(X_train , Y_train)
#import seaborn as sns
#sns.heatmap(X_train, annot=True)
from sklearn.metrics import plot_confusion_matrix
def conf_mat_acc(modelo):
disp = plot_confusion_matrix(modelo , X_train , Y_train ,
cmap = plt.cm.Blues , values_format = "d")
true_pred = disp.confusion_matrix[0,0]+disp.confusion_matrix[1,1]
total_data = np.sum(disp.confusion_matrix)
accuracy = true_pred/total_data
print("accuracy: " , np.round(accuracy , 2))
plt.show()
conf_mat_acc(decision_tree)
/shared-libs/python3.7/py/lib/python3.7/site-packages/sklearn/utils/deprecation.py:87: FutureWarning: Function plot_confusion_matrix is deprecated; Function `plot_confusion_matrix` is deprecated in 1.0 and will be removed in 1.2. Use one of the class methods: ConfusionMatrixDisplay.from_predictions or ConfusionMatrixDisplay.from_estimator.
warnings.warn(msg, category=FutureWarning)
accuracy: 1.0
# Preparar el test set para la evaluación
print(test.head())
test.info()
Make Model Nickname Car Gender Buyer Age \
0 Infiniti G Eliot Male 51
1 Chevrolet Suburban 2500 Bryna Male 59
2 Mitsubishi Precis Mack Female 32
3 Mercedes-Benz E-Class Elora Female 43
4 Plymouth Breeze Sigvard Male 34
Buzzword Country City Dealer Latitude \
0 multi-tasking Hungary Budapest 47.387906
1 framework Indonesia Liliba -10.170262
2 homogeneous China Shuangxing 41.773024
3 attitude-oriented France Grenoble 45.193486
4 Inverse Ukraine Khust 48.173463
Dealer Longitude Color New Car Purchase Date Sale Price Discount \
0 19.115039 Mauv VERDADERO 5/11/2015 70924.82 0.6790
1 123.642753 Crimson FALSO 2/04/2015 90407.26 0.1462
2 123.356112 Fuscia FALSO 5/06/2016 51744.01 0.1785
3 5.721898 Violet FALSO 29/05/2016 60654.60 0.2899
4 23.297248 Indigo FALSO 10/03/2009 68702.62 0.0721
Resell Price 5-yr Depreciation Top Speed 0-60 Time
0 13065.96 0.12 187.7 9.8
1 40998.62 0.16 197.9 2.5
2 17450.08 0.22 152.3 10.9
3 43694.28 0.06 179.0 5.6
4 19594.64 0.14 225.0 10.3
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5002 entries, 0 to 5001
Data columns (total 19 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Make 5002 non-null object
1 Model 5002 non-null object
2 Nickname 5002 non-null object
3 Car Gender 5002 non-null object
4 Buyer Age 5002 non-null int64
5 Buzzword 5002 non-null object
6 Country 5002 non-null object
7 City 5002 non-null object
8 Dealer Latitude 5002 non-null float64
9 Dealer Longitude 5002 non-null float64
10 Color 5002 non-null object
11 New Car 5002 non-null object
12 Purchase Date 5002 non-null object
13 Sale Price 5002 non-null float64
14 Discount 5002 non-null float64
15 Resell Price 5002 non-null float64
16 5-yr Depreciation 5002 non-null float64
17 Top Speed 5002 non-null float64
18 0-60 Time 5002 non-null float64
dtypes: float64(8), int64(1), object(10)
memory usage: 742.6+ KB
# Preparando test set
# Cambiando estado del carro
test['New Car'] = test["New Car"].map({"VERDADERO": 1, "FALSO": 0}).astype(int)
test.head(20)
Makeobject
Chevrolet15%
Dodge15%
12 others70%
Modelobject
G5%
Suburban 25005%
18 others90%
0
Infiniti
G
1
Chevrolet
Suburban 2500
2
Mitsubishi
Precis
3
Mercedes-Benz
E-Class
4
Plymouth
Breeze
5
Chevrolet
Avalanche
6
Dodge
Grand Caravan
7
Toyota
Celica
8
Spyker
C8 Double 12 S
9
GMC
Yukon XL 1500
print(test.info())
test[[
"Buyer Age",
"Top Speed",
"5-yr Depreciation",
"Sale Price"
]].head()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5002 entries, 0 to 5001
Data columns (total 19 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Make 5002 non-null object
1 Model 5002 non-null object
2 Nickname 5002 non-null object
3 Car Gender 5002 non-null object
4 Buyer Age 5002 non-null int64
5 Buzzword 5002 non-null object
6 Country 5002 non-null object
7 City 5002 non-null object
8 Dealer Latitude 5002 non-null float64
9 Dealer Longitude 5002 non-null float64
10 Color 5002 non-null object
11 New Car 5002 non-null int64
12 Purchase Date 5002 non-null object
13 Sale Price 5002 non-null float64
14 Discount 5002 non-null float64
15 Resell Price 5002 non-null float64
16 5-yr Depreciation 5002 non-null float64
17 Top Speed 5002 non-null float64
18 0-60 Time 5002 non-null float64
dtypes: float64(8), int64(2), object(9)
memory usage: 742.6+ KB
None
Buyer Ageint64
Top Speedfloat64
0
51
187.7
1
59
197.9
2
32
152.3
3
43
179
4
34
225
# Crear test set
X_test = test[[
"Buyer Age",
"Top Speed",
"5-yr Depreciation",
"Sale Price"
]]
print(X_test.shape)
from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_test = sc.fit_transform(X_test)
X_test
(5002, 4)
# Predicción de genero sobre el test set
Y_pred_tree = decision_tree.predict(X_test)
print("Clasificador arbol de decisión: ", Y_pred_tree[0 : 10])
Clasificador arbol de decisión: [1 0 1 0 1 1 1 1 1 0]
# Para guardar en ordenador
def download_output(y_pred, name):
output = pd.DataFrame({"Sale Price" : test['Sale Price'],
"Buyer Gender" : y_pred})
output.to_csv(name , index = False)
download_output(Y_pred_tree , "precio_genero_pred_tree.csv")
Conclusión
Durante el análisis de los datos del negocio, se obtuvo que el género femenino es el que más ingresos aporta a la concesionaria. Se obtiene que los ingresos por género son: Mujeres $210.040.800.000.000 Hombres $203.521.700.000.000
Luego, se utilizó el algoritmo de árbol de decisión, para crear un modelos y testearlo, de los cuales dieron como resultado de precisión del 100%. Por último se obtuvieron predicciones de un conjunto de datos de prueba con valores del 100% correctos sobre el género compraría un auto por su precio. El modelo predijo los siguientes resultados:
Clasificador árbol de decisión: [1 0 1 0 1 1 1 1 1 0 1 1 0 0 1 0 1 1 0 1]