Data Analysis with Feature Engineering Techniques

# Necessary Imports import numpy as np import pandas as pd pd.set_option("display.max_columns", None) import matplotlib.pyplot as plt %matplotlib inline plt.style.use("/home/lyrax/matplotlib-dracula/dracula.mplstyle") import seaborn as sns sns.set_palette("bright") from tqdm import tqdm, trange import warnings warnings.filterwarnings("ignore")

data = pd.read_csv("phonesdata.csv") data

# data info data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1148 entries, 0 to 1147
Data columns (total 7 columns):
 #   Column            Non-Null Count  Dtype  
---  ------            --------------  -----  
 0   Phone Title       1148 non-null   object 
 1   Specs             1148 non-null   object 
 2   Price(Kshs)       357 non-null    object 
 3   Rating            1148 non-null   float64
 4   Specs Score       1148 non-null   object 
 5   Likes             1148 non-null   int64  
 6   PriceJumia(Kshs)  1003 non-null   object 
dtypes: float64(1), int64(1), object(5)
memory usage: 62.9+ KB

# We'll use the PriceJumia column as the price column since it has few missing entries # Price column is the only with missing values so we may drop these few instances data['Price(Kshs)'] = data['PriceJumia(Kshs)'] data.drop('PriceJumia(Kshs)', axis=1, inplace=True) data.dropna(inplace=True); data.reset_index(drop=True, inplace=True)

def preprocess(df): # screen inches column df['screen_inches'] = df.Specs.apply(lambda spec: float(spec.split(',')[0].split()[0].split(':')[1])) # pixels column df['pixels'] = df.Specs.apply(lambda spec: spec.split(',')[1].partition('pixels')[0]) df['pixels_0'] = df.pixels.apply(lambda pix: int(pix.split()[0])) df['pixels_1'] = df.pixels.apply(lambda pix: int(pix.split()[-1])) # front cam df['front_cam(MP)'] = df.Specs.apply(lambda spec: spec.split(',')[2].partition('Front Camera:')[2].split('MP')[0]) # android df['android'] = np.nan for i, _ in enumerate(df.Specs): try: df.loc[i, 'android'] = df.loc[i, 'Specs'].split(',')[3].split()[1] except IndexError: pass df.android = df.android.apply(lambda s: str(s).split('.')[0]) df.android = df.android.str.replace(';','') for i, _ in enumerate(df.android): try: df.loc[i, 'android'] = float(df.loc[i, 'android']) except ValueError: df.loc[i, 'android'] = np.nan # storage df['storage(GB)'] = df.Specs.apply(lambda spec: spec.split(',')[4].partition('Storage:')[-1].split('GB')[0]) # ram df['ram(GB)'] = df.Specs.apply(lambda spec: spec.split(',')[5].partition('RAM:')[-1].split('GB')[0]) # battery df['battery(mAh)'] = df.Specs.apply(lambda spec: spec.split(',')[6].partition('Battery:')[-1].split('mAh')[0]) # Clean Other columns df['Price(Kshs)'] = df['Price(Kshs)'].apply(lambda pr: str(pr.split()[1])).str.replace(',', '').astype(int) df['Specs Score(%)'] = df['Specs Score'].str.replace('%', '').astype(int); df.drop('Specs Score', axis=1, inplace=True) # Applying the function preprocess(data) data

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1003 entries, 0 to 1002
Data columns (total 15 columns):
 #   Column          Non-Null Count  Dtype  
---  ------          --------------  -----  
 0   Phone Title     1003 non-null   object 
 1   Specs           1003 non-null   object 
 2   Price(Kshs)     1003 non-null   int64  
 3   Rating          1003 non-null   float64
 4   Likes           1003 non-null   int64  
 5   screen_inches   1003 non-null   float64
 6   pixels          1003 non-null   object 
 7   pixels_0        1003 non-null   int64  
 8   pixels_1        1003 non-null   int64  
 9   front_cam(MP)   1003 non-null   object 
 10  android         873 non-null    object 
 11  storage(GB)     1003 non-null   object 
 12  ram(GB)         1003 non-null   object 
 13  battery(mAh)    1003 non-null   object 
 14  Specs Score(%)  1003 non-null   int64  
dtypes: float64(2), int64(5), object(8)
memory usage: 117.7+ KB

# drop instances without android version data.dropna(inplace=True); data.reset_index(drop=True, inplace=True) data

plt.figure(figsize=(9,6)) sns.distplot(data['Price(Kshs)'])

print(f"Minimum price: {data['Price(Kshs)'].min()} KShs.\nMaximum price: {data['Price(Kshs)'].max()} KShs.")

Minimum price: 124 KShs.
Maximum price: 446342 KShs.

data_clipped = data[data['Price(Kshs)'] == np.clip(data['Price(Kshs)'], 4000, 150000)] data_clipped

# replotting price distribution plt.figure(figsize=(9,6)) sns.distplot(data_clipped['Price(Kshs)'])

# distribution and boxplot of phone Rating fig, ax = plt.subplots(1, 2, figsize=(17, 6)) sns.distplot(data_clipped.Rating, ax=ax[0]); ax[0].set_title("Distribution of phone Ratings") sns.boxplot(data_clipped.Rating, ax=ax[1]); ax[1].set_title("Boxplot")

plt.figure(figsize=(12,7)) sns.scatterplot(data_clipped['Price(Kshs)'], data_clipped.Rating) sns.jointplot(data_clipped['Price(Kshs)'], data_clipped.Rating)

plt.figure(figsize=(12,7)) sns.scatterplot(data_clipped.Rating, data_clipped.Likes) sns.jointplot(data_clipped.Rating, data_clipped.Likes)

plt.figure(figsize=(8,15)) sns.countplot(y=data_clipped.screen_inches)

plt.figure(figsize=(12,7)) sns.countplot(data_clipped.android)

data_clipped = data_clipped[data_clipped.android <= 10]

plt.figure(figsize=(12,7)) sns.scatterplot(data_clipped.Rating, data_clipped['Specs Score(%)']) sns.jointplot(data_clipped.Rating, data_clipped['Specs Score(%)'])

plt.figure(figsize=(12,7)) sns.scatterplot(data_clipped.Likes, data_clipped['Specs Score(%)']) sns.jointplot(data_clipped.Likes, data_clipped['Specs Score(%)'])