In [ ]:
Objective and Focus:
Our primary goal is to enhance marketing strategies among existing clients who are not currently making purchases, aiming to stimulate their interest in buying. This initiative emphasizes addressing false positives to effectively convert potential clients.
In contrast, if our aim were to encourage repeat purchases from existing buyers, our strategy would shift. We would focus on understanding:
- The demographic profile of non-purchasing clients.
- Identifying reasons behind their lack of purchases:
- Are they satisfied with another brand?
- Are financial constraints influencing their decision? Insights from their salary can provide clarity.
- Are they occasional buyers who only purchase when there's a specific need?
In [317]:
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
In [318]:
df = pd.read_csv(r"C:\Users\Teni\Desktop\Git-Github\Datasets\Logistic Regression\car_data.csv")
In [319]:
df
Out[319]:
| User ID | Gender | Age | AnnualSalary | Purchased | |
|---|---|---|---|---|---|
| 0 | 385 | Male | 35 | 20000 | 0 |
| 1 | 681 | Male | 40 | 43500 | 0 |
| 2 | 353 | Male | 49 | 74000 | 0 |
| 3 | 895 | Male | 40 | 107500 | 1 |
| 4 | 661 | Male | 25 | 79000 | 0 |
| ... | ... | ... | ... | ... | ... |
| 995 | 863 | Male | 38 | 59000 | 0 |
| 996 | 800 | Female | 47 | 23500 | 0 |
| 997 | 407 | Female | 28 | 138500 | 1 |
| 998 | 299 | Female | 48 | 134000 | 1 |
| 999 | 687 | Female | 44 | 73500 | 0 |
1000 rows × 5 columns
In [320]:
df.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 1000 entries, 0 to 999 Data columns (total 5 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 User ID 1000 non-null int64 1 Gender 1000 non-null object 2 Age 1000 non-null int64 3 AnnualSalary 1000 non-null int64 4 Purchased 1000 non-null int64 dtypes: int64(4), object(1) memory usage: 39.2+ KB
General Observation about the data
- the data has 1000 rows
- the lowest age surveyed is 18 and the maximum ae is 63, the mean is 40
- the min salary is 15000, while the maximum salary is 152500
- the std of the age is low (10) meaning the majority of the data is centreed around the 30 (mean - std)30 & 50 (mean + the std). Approximately 68% of the data will fall within the range of 30-50
Data Cleaning¶
In [321]:
# take the User ID, since we have our unique IDs
df.drop('User ID', axis=1, inplace=True)
df
Out[321]:
| Gender | Age | AnnualSalary | Purchased | |
|---|---|---|---|---|
| 0 | Male | 35 | 20000 | 0 |
| 1 | Male | 40 | 43500 | 0 |
| 2 | Male | 49 | 74000 | 0 |
| 3 | Male | 40 | 107500 | 1 |
| 4 | Male | 25 | 79000 | 0 |
| ... | ... | ... | ... | ... |
| 995 | Male | 38 | 59000 | 0 |
| 996 | Female | 47 | 23500 | 0 |
| 997 | Female | 28 | 138500 | 1 |
| 998 | Female | 48 | 134000 | 1 |
| 999 | Female | 44 | 73500 | 0 |
1000 rows × 4 columns
In [322]:
# convert gender to a dummy variable
df = pd.get_dummies(df, columns=['Gender'])
df
Out[322]:
| Age | AnnualSalary | Purchased | Gender_Female | Gender_Male | |
|---|---|---|---|---|---|
| 0 | 35 | 20000 | 0 | 0 | 1 |
| 1 | 40 | 43500 | 0 | 0 | 1 |
| 2 | 49 | 74000 | 0 | 0 | 1 |
| 3 | 40 | 107500 | 1 | 0 | 1 |
| 4 | 25 | 79000 | 0 | 0 | 1 |
| ... | ... | ... | ... | ... | ... |
| 995 | 38 | 59000 | 0 | 0 | 1 |
| 996 | 47 | 23500 | 0 | 1 | 0 |
| 997 | 28 | 138500 | 1 | 1 | 0 |
| 998 | 48 | 134000 | 1 | 1 | 0 |
| 999 | 44 | 73500 | 0 | 1 | 0 |
1000 rows × 5 columns
In [323]:
df = df.drop('Gender_Male', axis=1)
df
Out[323]:
| Age | AnnualSalary | Purchased | Gender_Female | |
|---|---|---|---|---|
| 0 | 35 | 20000 | 0 | 0 |
| 1 | 40 | 43500 | 0 | 0 |
| 2 | 49 | 74000 | 0 | 0 |
| 3 | 40 | 107500 | 1 | 0 |
| 4 | 25 | 79000 | 0 | 0 |
| ... | ... | ... | ... | ... |
| 995 | 38 | 59000 | 0 | 0 |
| 996 | 47 | 23500 | 0 | 1 |
| 997 | 28 | 138500 | 1 | 1 |
| 998 | 48 | 134000 | 1 | 1 |
| 999 | 44 | 73500 | 0 | 1 |
1000 rows × 4 columns
In [324]:
df = df.rename(columns={'Gender_Female':'female_gender'})
df
Out[324]:
| Age | AnnualSalary | Purchased | female_gender | |
|---|---|---|---|---|
| 0 | 35 | 20000 | 0 | 0 |
| 1 | 40 | 43500 | 0 | 0 |
| 2 | 49 | 74000 | 0 | 0 |
| 3 | 40 | 107500 | 1 | 0 |
| 4 | 25 | 79000 | 0 | 0 |
| ... | ... | ... | ... | ... |
| 995 | 38 | 59000 | 0 | 0 |
| 996 | 47 | 23500 | 0 | 1 |
| 997 | 28 | 138500 | 1 | 1 |
| 998 | 48 | 134000 | 1 | 1 |
| 999 | 44 | 73500 | 0 | 1 |
1000 rows × 4 columns
In [325]:
df.describe()
Out[325]:
| Age | AnnualSalary | Purchased | female_gender | |
|---|---|---|---|---|
| count | 1000.000000 | 1000.000000 | 1000.000000 | 1000.000000 |
| mean | 40.106000 | 72689.000000 | 0.402000 | 0.516000 |
| std | 10.707073 | 34488.341867 | 0.490547 | 0.499994 |
| min | 18.000000 | 15000.000000 | 0.000000 | 0.000000 |
| 25% | 32.000000 | 46375.000000 | 0.000000 | 0.000000 |
| 50% | 40.000000 | 72000.000000 | 0.000000 | 1.000000 |
| 75% | 48.000000 | 90000.000000 | 1.000000 | 1.000000 |
| max | 63.000000 | 152500.000000 | 1.000000 | 1.000000 |
In [326]:
# checking for null data.
df.isnull().sum()
Out[326]:
Age 0 AnnualSalary 0 Purchased 0 female_gender 0 dtype: int64
Exploratory Data Analysis (EDA)¶
age and purchaese, gender-f,
In [327]:
# checjk if the data is ballanced
df['Purchased'].value_counts()
Out[327]:
0 598 1 402 Name: Purchased, dtype: int64
In [328]:
sns.countplot(data=df, x='Purchased')
plt.show()
In [329]:
plt.figure(figsize=(12, 8), dpi=200)
sns.scatterplot(data=df, x='Age', y='AnnualSalary', hue='Purchased')
plt.tight_layout()
plt.show()
In [330]:
plt.figure(figsize=(8,6), dpi=200)
sns.scatterplot(data=df, x='Age', y='AnnualSalary', hue='female_gender')
# plt.tight_layout()
plt.show()
In [331]:
sns.countplot(x='female_gender',data=df, hue='Purchased')
plt.title('Purchases for the Male (0) and the Female(1) Gender')
plt.legend(title='Purchased', loc='upper right')
plt.show()
Notes from the above:
- Female participants in this survey purchased more cars than their male counterparts.
- The older generation showed a higher car purchase rate, indicating changing priorities, even if their income wasn't significantly higher.
- It's likely that the younger demographic has other priorities such as travel, education, personal development, exams, or starting a family.
In [332]:
sns.heatmap(df.corr(), annot=True)
plt.show()
Notes from the analysis
- annual salary has the greatest correlation with if a car is purchased or not
- as the age progresses, annula salary decrease
Define the X and Y¶
In [333]:
X = df.drop('Purchased', axis=1)
y = df.Purchased
Train Test Split¶
In [334]:
from sklearn.model_selection import train_test_split
In [335]:
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.30, random_state=101)
Scale the data¶
In [336]:
from sklearn.preprocessing import StandardScaler
In [337]:
scaler = StandardScaler()
In [338]:
scaled_x_train = scaler.fit_transform(X_train)
In [339]:
scaled_x_test = scaler.transform(X_test)
Create the model and predict¶
In [340]:
from sklearn.linear_model import LogisticRegression
In [341]:
model = LogisticRegression()
Predict¶
In [342]:
model.fit(scaled_x_train, y_train)
Out[342]:
LogisticRegression()
In [343]:
y_pred= model.predict(scaled_x_test)
y_pred
Out[343]:
array([0, 1, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 0, 0, 0,
1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 1,
0, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0,
1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 0, 0,
0, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0,
0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0,
0, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0,
1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,
0, 0, 1, 1, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 1, 0,
0, 0, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1,
0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, 1, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 1,
0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 0, 1, 1,
0, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 0, 0], dtype=int64)
In [344]:
model.predict_proba(scaled_x_test)
Out[344]:
array([[0.99868814, 0.00131186],
[0.4578815 , 0.5421185 ],
[0.21962047, 0.78037953],
[0.67531156, 0.32468844],
[0.41424985, 0.58575015],
[0.58345303, 0.41654697],
[0.88087378, 0.11912622],
[0.91304898, 0.08695102],
[0.0363067 , 0.9636933 ],
[0.68232584, 0.31767416],
[0.70901873, 0.29098127],
[0.99010544, 0.00989456],
[0.30142507, 0.69857493],
[0.41049669, 0.58950331],
[0.10548753, 0.89451247],
[0.99798686, 0.00201314],
[0.62449971, 0.37550029],
[0.04277203, 0.95722797],
[0.86710804, 0.13289196],
[0.94882396, 0.05117604],
[0.85168651, 0.14831349],
[0.93124916, 0.06875084],
[0.45823822, 0.54176178],
[0.94570206, 0.05429794],
[0.92132013, 0.07867987],
[0.58310375, 0.41689625],
[0.21186749, 0.78813251],
[0.62189261, 0.37810739],
[0.96706985, 0.03293015],
[0.97900077, 0.02099923],
[0.86897349, 0.13102651],
[0.97264627, 0.02735373],
[0.93593479, 0.06406521],
[0.50887591, 0.49112409],
[0.71231788, 0.28768212],
[0.99009763, 0.00990237],
[0.84136965, 0.15863035],
[0.74075443, 0.25924557],
[0.18373342, 0.81626658],
[0.02972873, 0.97027127],
[0.88915712, 0.11084288],
[0.758928 , 0.241072 ],
[0.00967984, 0.99032016],
[0.06732073, 0.93267927],
[0.56011819, 0.43988181],
[0.00178187, 0.99821813],
[0.78585194, 0.21414806],
[0.97867166, 0.02132834],
[0.98893697, 0.01106303],
[0.91309316, 0.08690684],
[0.07234919, 0.92765081],
[0.89958057, 0.10041943],
[0.00357407, 0.99642593],
[0.08628479, 0.91371521],
[0.59483428, 0.40516572],
[0.55679062, 0.44320938],
[0.97865501, 0.02134499],
[0.79190042, 0.20809958],
[0.99240978, 0.00759022],
[0.96819729, 0.03180271],
[0.68280998, 0.31719002],
[0.09981866, 0.90018134],
[0.81720505, 0.18279495],
[0.11304552, 0.88695448],
[0.92793174, 0.07206826],
[0.99602699, 0.00397301],
[0.15723429, 0.84276571],
[0.01546607, 0.98453393],
[0.16648082, 0.83351918],
[0.02747839, 0.97252161],
[0.95257535, 0.04742465],
[0.27591275, 0.72408725],
[0.47325596, 0.52674404],
[0.94251796, 0.05748204],
[0.68567355, 0.31432645],
[0.95399031, 0.04600969],
[0.98509682, 0.01490318],
[0.99644734, 0.00355266],
[0.96176422, 0.03823578],
[0.25148116, 0.74851884],
[0.4648858 , 0.5351142 ],
[0.7970236 , 0.2029764 ],
[0.03559032, 0.96440968],
[0.78648175, 0.21351825],
[0.93970015, 0.06029985],
[0.0077571 , 0.9922429 ],
[0.99309785, 0.00690215],
[0.99418662, 0.00581338],
[0.77269891, 0.22730109],
[0.98664968, 0.01335032],
[0.90637647, 0.09362353],
[0.04500582, 0.95499418],
[0.4375525 , 0.5624475 ],
[0.74351973, 0.25648027],
[0.6954827 , 0.3045173 ],
[0.07037575, 0.92962425],
[0.91892638, 0.08107362],
[0.36095152, 0.63904848],
[0.04676752, 0.95323248],
[0.81418103, 0.18581897],
[0.9876371 , 0.0123629 ],
[0.91166373, 0.08833627],
[0.9681038 , 0.0318962 ],
[0.74389973, 0.25610027],
[0.24018896, 0.75981104],
[0.09799778, 0.90200222],
[0.54804731, 0.45195269],
[0.92240704, 0.07759296],
[0.13361612, 0.86638388],
[0.71195855, 0.28804145],
[0.54440597, 0.45559403],
[0.76196736, 0.23803264],
[0.65851906, 0.34148094],
[0.95607775, 0.04392225],
[0.00269075, 0.99730925],
[0.97581327, 0.02418673],
[0.11672222, 0.88327778],
[0.7919401 , 0.2080599 ],
[0.74960686, 0.25039314],
[0.37579215, 0.62420785],
[0.58721234, 0.41278766],
[0.50550292, 0.49449708],
[0.99866716, 0.00133284],
[0.94500493, 0.05499507],
[0.36924082, 0.63075918],
[0.98840042, 0.01159958],
[0.99556454, 0.00443546],
[0.32937099, 0.67062901],
[0.51224809, 0.48775191],
[0.31837644, 0.68162356],
[0.61052723, 0.38947277],
[0.94962052, 0.05037948],
[0.93109583, 0.06890417],
[0.7892439 , 0.2107561 ],
[0.80453756, 0.19546244],
[0.00704674, 0.99295326],
[0.72571862, 0.27428138],
[0.01062508, 0.98937492],
[0.00601307, 0.99398693],
[0.78438248, 0.21561752],
[0.89347987, 0.10652013],
[0.34307197, 0.65692803],
[0.11676333, 0.88323667],
[0.4895573 , 0.5104427 ],
[0.68241226, 0.31758774],
[0.84966846, 0.15033154],
[0.66136856, 0.33863144],
[0.99666057, 0.00333943],
[0.17675084, 0.82324916],
[0.98910685, 0.01089315],
[0.95180485, 0.04819515],
[0.73471456, 0.26528544],
[0.95185246, 0.04814754],
[0.85171692, 0.14828308],
[0.11151748, 0.88848252],
[0.92151079, 0.07848921],
[0.86519206, 0.13480794],
[0.78648175, 0.21351825],
[0.99454823, 0.00545177],
[0.01406308, 0.98593692],
[0.77276893, 0.22723107],
[0.91181145, 0.08818855],
[0.74699006, 0.25300994],
[0.73838565, 0.26161435],
[0.64417965, 0.35582035],
[0.9988068 , 0.0011932 ],
[0.97586028, 0.02413972],
[0.50887591, 0.49112409],
[0.59173234, 0.40826766],
[0.96905381, 0.03094619],
[0.89067443, 0.10932557],
[0.9592641 , 0.0407359 ],
[0.66468612, 0.33531388],
[0.01501908, 0.98498092],
[0.80963328, 0.19036672],
[0.93109583, 0.06890417],
[0.55659382, 0.44340618],
[0.56318497, 0.43681503],
[0.20473145, 0.79526855],
[0.08495971, 0.91504029],
[0.01673647, 0.98326353],
[0.09383084, 0.90616916],
[0.69979236, 0.30020764],
[0.08973859, 0.91026141],
[0.74402155, 0.25597845],
[0.00531604, 0.99468396],
[0.27890586, 0.72109414],
[0.96667084, 0.03332916],
[0.86680432, 0.13319568],
[0.04842461, 0.95157539],
[0.93010575, 0.06989425],
[0.04543352, 0.95456648],
[0.91287142, 0.08712858],
[0.99097701, 0.00902299],
[0.57535981, 0.42464019],
[0.76742593, 0.23257407],
[0.46498499, 0.53501501],
[0.79448061, 0.20551939],
[0.94891302, 0.05108698],
[0.94802736, 0.05197264],
[0.9481647 , 0.0518353 ],
[0.05211272, 0.94788728],
[0.20988726, 0.79011274],
[0.02213708, 0.97786292],
[0.01200981, 0.98799019],
[0.03471552, 0.96528448],
[0.9902383 , 0.0097617 ],
[0.35793624, 0.64206376],
[0.20676704, 0.79323296],
[0.12999832, 0.87000168],
[0.72540104, 0.27459896],
[0.97696692, 0.02303308],
[0.93104465, 0.06895535],
[0.78900478, 0.21099522],
[0.02111203, 0.97888797],
[0.01014198, 0.98985802],
[0.04351013, 0.95648987],
[0.07008934, 0.92991066],
[0.52111817, 0.47888183],
[0.36224013, 0.63775987],
[0.88267115, 0.11732885],
[0.85578363, 0.14421637],
[0.29173214, 0.70826786],
[0.64702969, 0.35297031],
[0.93704567, 0.06295433],
[0.99455341, 0.00544659],
[0.50927451, 0.49072549],
[0.5243854 , 0.4756146 ],
[0.00652919, 0.99347081],
[0.9841183 , 0.0158817 ],
[0.72540104, 0.27459896],
[0.64353964, 0.35646036],
[0.1279004 , 0.8720996 ],
[0.8598533 , 0.1401467 ],
[0.33923222, 0.66076778],
[0.56783351, 0.43216649],
[0.95269043, 0.04730957],
[0.943125 , 0.056875 ],
[0.31155501, 0.68844499],
[0.45685274, 0.54314726],
[0.83937776, 0.16062224],
[0.60986336, 0.39013664],
[0.92908173, 0.07091827],
[0.76454987, 0.23545013],
[0.99286703, 0.00713297],
[0.92488957, 0.07511043],
[0.01061923, 0.98938077],
[0.99670649, 0.00329351],
[0.95859128, 0.04140872],
[0.99299087, 0.00700913],
[0.98907243, 0.01092757],
[0.07451091, 0.92548909],
[0.38084153, 0.61915847],
[0.75284446, 0.24715554],
[0.87241597, 0.12758403],
[0.97935231, 0.02064769],
[0.38845651, 0.61154349],
[0.88594632, 0.11405368],
[0.22606228, 0.77393772],
[0.61764172, 0.38235828],
[0.87238916, 0.12761084],
[0.65133835, 0.34866165],
[0.73176305, 0.26823695],
[0.0447665 , 0.9552335 ],
[0.7913482 , 0.2086518 ],
[0.82837723, 0.17162277],
[0.9398445 , 0.0601555 ],
[0.98727316, 0.01272684],
[0.95408465, 0.04591535],
[0.71870433, 0.28129567],
[0.24914362, 0.75085638],
[0.81684741, 0.18315259],
[0.97617055, 0.02382945],
[0.7403251 , 0.2596749 ],
[0.00934296, 0.99065704],
[0.04219068, 0.95780932],
[0.15110327, 0.84889673],
[0.42129957, 0.57870043],
[0.38112363, 0.61887637],
[0.67553917, 0.32446083],
[0.79461079, 0.20538921],
[0.65088547, 0.34911453],
[0.22899472, 0.77100528],
[0.52052112, 0.47947888],
[0.10840788, 0.89159212],
[0.06739587, 0.93260413],
[0.79418071, 0.20581929],
[0.81911756, 0.18088244],
[0.94147565, 0.05852435],
[0.94570206, 0.05429794],
[0.40298395, 0.59701605],
[0.00907348, 0.99092652],
[0.43335287, 0.56664713],
[0.64262446, 0.35737554],
[0.05696321, 0.94303679],
[0.61444976, 0.38555024],
[0.92483415, 0.07516585],
[0.36238789, 0.63761211],
[0.92155105, 0.07844895],
[0.60632465, 0.39367535]])
Model Evaluation¶
In [345]:
from sklearn.metrics import accuracy_score, classification_report, confusion_matrix, plot_confusion_matrix
In [346]:
accuracy_score(y_pred, y_test)
# the truly predicted values of the model make up 84% of the matrix
Out[346]:
0.8366666666666667
In [347]:
confusion_matrix(y_test, y_pred)
Out[347]:
array([[167, 19],
[ 30, 84]], dtype=int64)
In [348]:
plot_confusion_matrix(model, scaled_x_test, y_test, cmap=plt.cm.Blues)
C:\Users\Teni\anaconda3\lib\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)
Out[348]:
<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x1af9e493640>
Notes:
- 167 and 84 are the true values for negatives and positives, respectively.
- 30 is the False Negative, while 19 is the False Positive.
- Since the objective of our prediction is to improve marketing strategies for existing clients who don’t purchase, the focus will be on minimizing both false positives and false negatives. This ensures we accurately identify potential clients to convert.
- If the focus was on stimulating those who already buy to keep buying, our approach would differ.
- To avoid missing out on important clients or data points, we'll hyper-tune our model to minimize and ensure the reliability of both false positives and false negatives.
- Therefore, the focus will be on Recall Metrics because the cost of missing a False Positive is high.
In [301]:
print(classification_report(y_test, y_pred))
precision recall f1-score support
0 0.85 0.90 0.87 186
1 0.82 0.74 0.77 114
accuracy 0.84 300
macro avg 0.83 0.82 0.82 300
weighted avg 0.84 0.84 0.83 300
- Need to focus on minimizing the False Positives so that we don't miss on potential clients we can market to due to the model's prediction....
- And for that, focus will be on optimizing the precision Score
Hypertune the Model for a better Precision Score¶
Using GridSearchCV on the Logistic Regression Model¶
In [304]:
from sklearn.linear_model import LogisticRegression
lr_model = LogisticRegression(max_iter=100)
lr_model.fit(scaled_x_train, y_train)
Out[304]:
LogisticRegression()
In [305]:
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import make_scorer, precision_score
from sklearn.preprocessing import StandardScaler
param_grid = {
'C': [0.1, 1, 10, 100],
'solver': [ 'liblinear'],
'penalty': ['l1', 'l2'],
}
scaler = StandardScaler()
scaled_x_train = scaler.fit_transform(X_train)
scaled_x_test = scaler.transform(X_test)
scorer = make_scorer(precision_score, average='weighted')
grid_search = GridSearchCV(lr_model, param_grid, cv=7, scoring=scorer)
grid_search.fit(scaled_x_train, y_train)
print("Best hyperparameters: ", grid_search.best_params_)
print("Best cross_validation precision score: ",grid_search.best_score_)
Best hyperparameters: {'C': 0.1, 'penalty': 'l1', 'solver': 'liblinear'}
Best cross_validation precision score: 0.8392065150308922
In [306]:
lr_model.fit(scaled_x_train, y_train)
y_pred = lr_model.predict(scaled_x_test)
In [307]:
print(classification_report(y_test, y_pred))
precision recall f1-score support
0 0.85 0.90 0.87 186
1 0.82 0.74 0.77 114
accuracy 0.84 300
macro avg 0.83 0.82 0.82 300
weighted avg 0.84 0.84 0.83 300
Got the same result with GridSearchV. So, a different model will be used
Adjusting the model- using RandomForestClassifier¶
In [308]:
from sklearn.ensemble import RandomForestClassifier
# Example of using Random Forest for potentially better performance
rf_model = RandomForestClassifier(n_estimators=100, random_state=42)
scaler = StandardScaler()
rf_scaled_x_train = scaler.fit_transform(X_train)
rf_scaled_x_test = scaler.transform(X_test)
rf_model.fit(rf_scaled_x_train, y_train)
y_pred_rf = rf_model.predict(rf_scaled_x_test)
# Evaluate performance metrics
from sklearn.metrics import classification_report
print("Random Forest Classification Report:")
print(classification_report(y_test, y_pred_rf))
Random Forest Classification Report:
precision recall f1-score support
0 0.92 0.92 0.92 186
1 0.87 0.88 0.87 114
accuracy 0.90 300
macro avg 0.90 0.90 0.90 300
weighted avg 0.90 0.90 0.90 300
In [309]:
plot_confusion_matrix(rf_model, scaled_x_test, y_test)
C:\Users\Teni\anaconda3\lib\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)
Out[309]:
<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x1af9d7f6640>
Note
- the RandomForestClassifier model is the most viable model for this data
In [310]:
confusion_matrix(y_test, y_pred_rf)
Out[310]:
array([[171, 15],
[ 14, 100]], dtype=int64)
To Deploy¶
In [311]:
final_model = RandomForestClassifier(n_estimators=100, random_state=42)
final_model.fit(X, y)
y_pred = final_model.predict(X)
In [312]:
from joblib import dump, load
dump(final_model, 'car_sales_pred.joblib')
Out[312]:
['car_sales_pred.joblib']
In [313]:
model = load('car_sales_pred.joblib')
Conclusion and Recommendations for Marketers:
Key Observations:
- Female participants in our survey demonstrated a higher rate of car purchases compared to males.
- The older generation exhibited a higher propensity to purchase cars, suggesting shifting priorities even without significantly higher incomes.
- Younger demographics likely prioritize other life aspects such as travel, education, personal growth, exams, or family planning.
Insights for Marketers:
- Annual salary shows the strongest correlation with car purchase behavior.
- As age increases, annual salary tends to decrease.
Recommendations:
Targeting Strategies: Develop targeted campaigns that resonate with the preferences and priorities of different demographic groups. For instance, focus on convenience and value propositions for older buyers, while emphasizing lifestyle benefits for younger demographics.
Financial Accessibility: Consider flexible financing options or promotional offers that address potential financial constraints identified among non-purchasing clients.
Retention Strategies: Implement loyalty programs and personalized marketing to retain existing clients and encourage repeat purchases.
In [ ]:
In [ ]: