In [1]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from matplotlib import gridspec
import statsmodels.api as sm
In [10]:
#Importing data
data = pd.read_csv("Boats.csv")
In [11]:
data.head()
Out[11]:
Unnamed: 0 id type boatClass make model year condition length_ft beam_ft ... minEngineYear engineCategory price sellerId city state zip created_date created_month created_year
0 1 7252689 power power-center Aquasport 210 CC 1992 used 21.0 8.50 ... NaN outboard-4s 16500.0 217053 Englewood FL 34224 2019-10-16 10 2019
1 3 7228300 power power-sportcruiser Formula 400 Super Sport 2018 used 40.0 11.00 ... 2018.0 inboard-outboard 539000.0 44260 Harsens Island MI 48028 2019-09-24 9 2019
2 5 7271336 power power-deck Bayliner Element 180 2020 new 18.0 7.42 ... 2019.0 outboard-4s 26995.0 220570 Marietta OH 45750 2019-11-02 11 2019
3 6 7222952 power power-expresscruiser Regal 32 Express 2015 used 32.0 10.33 ... NaN multiple 169995.0 34834 North Charleston SC 29405 2019-09-19 9 2019
4 8 6824832 power power-aft Carver 440 Aft Cabin Motor Yacht 1994 used 44.0 15.00 ... 1994.0 inboard 109900.0 17942 Middle River MD 21220 2018-08-29 8 2018

5 rows × 26 columns

In [6]:
# Take selected columns
data.columns
data = data[['type','boatClass','make','model','year','condition','length_ft','beam_ft','dryWeight_lb','hullMaterial','engineCategory','price','city','state','zip','created_date']]

#Removing Null Values 
data = data.dropna()
In [13]:
# Data Overview
print("Rows",data.shape[0])
print("Columns",data.shape[1])
print("Missing values",data.isna().sum().values.sum())
print("Unique Values",data.nunique())

Rows 18903
Columns 26
Missing values 74812
Unique Values Unnamed: 0        18903
id                18903
type                  3
boatClass            72
make                960
model              7899
year                 80
condition             2
length_ft           599
beam_ft             350
dryWeight_lb       1160
hullMaterial          9
fuelType              4
numEngines            5
totalHP             286
maxEngineYear        67
minEngineYear        66
engineCategory        9
price              6074
sellerId           1375
city               1149
state                49
zip                1046
created_date       1424
created_month        12
created_year         16
dtype: int64
In [14]:
def value_count(column):
    plt.figure(figsize=(10,5))
    plt.grid(True)
    column.value_counts().plot(kind ="barh")
In [15]:
data.columns
value_count(data["condition"])
value_count(data["engineCategory"])
value_count(data["hullMaterial"])
In [16]:
data.describe()
Out[16]:
Unnamed: 0 id year length_ft beam_ft dryWeight_lb numEngines totalHP maxEngineYear minEngineYear price sellerId created_month created_year
count 18903.000000 1.890300e+04 18903.000000 18903.000000 12399.00000 7094.000000 18903.000000 18055.000000 2205.000000 2174.000000 1.890300e+04 18903.000000 18903.000000 18903.000000
mean 9782.042216 6.947263e+06 2013.145956 23.803935 16.19693 4754.142697 1.069513 112.849006 2007.927438 2007.919503 6.471469e+05 49891.652225 6.945511 2018.503465
std 5749.171393 4.820519e+05 10.502989 14.613329 36.83357 11626.382266 0.425528 286.137591 12.429153 12.462496 7.309567e+07 60725.592092 3.007115 1.093103
min 1.000000 4.449130e+05 1910.000000 1.000000 0.08000 8.000000 0.000000 0.000000 1938.000000 1938.000000 5.000000e+02 1003.000000 1.000000 2003.000000
25% 4783.500000 6.895201e+06 2011.000000 18.000000 7.83000 1175.000000 1.000000 0.000000 2001.000000 2001.000000 1.925500e+04 10550.000000 5.000000 2018.000000
50% 9603.000000 7.061616e+06 2019.000000 21.000000 8.50000 2001.000000 1.000000 0.000000 2012.000000 2012.000000 3.419500e+04 34482.000000 8.000000 2019.000000
75% 14718.500000 7.179152e+06 2019.000000 25.000000 9.00000 3375.000000 1.000000 115.000000 2019.000000 2019.000000 5.783000e+04 53226.000000 9.000000 2019.000000
max 20000.000000 7.271336e+06 2020.000000 375.000000 1311.00000 440000.000000 4.000000 7200.000000 2020.000000 2020.000000 1.000000e+10 269557.000000 12.000000 2019.000000
In [54]:
# length in feets , beam in feats, dryweight, totalHP, price are continues variable
In [17]:
#Exploring Price
def plot_3_ways(df,feature):
    fig = plt.figure(constrained_layout = True, figsize=(12,8))
    grid = gridspec.GridSpec(ncols =3, nrows =3, figure = fig)
    
    ax1 = fig.add_subplot(grid[0:2])
    ax1.set_title("Histogram")
    
    sns.distplot(df.loc[:,feature])
    plt.axvline(x = df[feature].mean(), c = 'red')
    plt.axvline(x = df[feature].median(), c = 'green')
    
    ax2 = fig.add_subplot(grid[1:2])
    sm.qqplot(df[feature])
    
    ax3 = fig.add_subplot(grid[:, 2])
    ax3.set_title("boxplot")
    sns.boxplot(df[feature], ax=ax3, orient ="v")
    
plot_3_ways(data,"price")

It shows our distribution is heavily skewed to the right side.

In [19]:
#changing price to log price and observing the new distribution
data["log_price"] = np.log(data["price"])
plot_3_ways(data,"log_price")

on transforming price to log, it gives us a better distribution of price.

In [20]:
# Let's observe the effect of price on new and used boats.
plt.figure(figsize=(10,5))
ax = sns.violinplot(x="condition", y="log_price", data=data)
In [21]:
#Creating a new variable age
data["age"] = data["created_year"] - data["year"]
#grid = gridspec.GridSpec(ncols =3, nrows =3, figure = fig)
    
#ax1 = fig.add_subplot(grid[0:1])
#ax1.set_title("Age vs Price")
#plt.axis([0, 100, 0, 0.2])
plt.figure(figsize=(10,5))
scatter = sns.scatterplot(data=data, x='age', y='log_price',hue ="condition")
#scatter.set_ylim(bottom=0, top=0.6)
#sns.scatterplot(data=data, x="age", y="log_price", hue='condition')
In [142]:
plt.figure(figsize=(10,5))
scatter = sns.scatterplot(data=data, x='age', y='log_price',hue ="type")
In [150]:
#sns.set(rc={'figure.figsize':(20,10)})
sns.catplot(x="engineCategory", y="log_price", hue="condition", kind="box", data=data,height=8.27, aspect=11.7/8.27)
Out[150]:
<seaborn.axisgrid.FacetGrid at 0x1b48b9a5888>

Observing Correlation ofcontinues variables in order to look for any correlation. relationship of price with length, beam_ft, age, dryweight_lb,totalHP

In [28]:
numerical_features = ['length_ft','beam_ft','dryWeight_lb','log_price','age','totalHP']
data_numerical = data[numerical_features]
sns.pairplot(data_numerical)
Out[28]:
<seaborn.axisgrid.PairGrid at 0x24f0df7b188>

There is definitely some relationship of price with age, length, dryweight and totalHP.

In [ ]:
#watching the relationship between price,length and boat_class
In [25]:
def top_5(data,column1_category,column2_continues,hue):
    new_df = data[column1_category].value_counts().nlargest(5).reset_index()
    series = new_df["index"].tolist()
    df_filtered = data[data[column1_category].isin(series)]
    plt.figure(figsize=(10,10))
    ax = sns.boxplot(y=column2_continues, x=column1_category, hue=hue,
                 data=df_filtered)
    return series
    

top_5(data,'boatClass','log_price','condition')
Out[25]:
['power-pontoon',
 'power-bowrider',
 'power-bass',
 'power-center',
 'power-cruiser']
In [26]:
top_5(data,'engineCategory','log_price','condition')
Out[26]:
['outboard', 'inboard', 'inboard-outboard', 'outboard-4s', 'other']

seems like older boats have higher prices for the engine category like outboard, inboard-outboard and inboard.

In [51]:
#Horse Power and price with new and old boat
#fueltype and horsePower
#Does Dryweight has anthing to do with new and old boats and if it does how does it affect price overall
#Age, Engine, Price
#Engine and Price
#Geographical Plots.
#summary
Out[51]:
['power-pontoon',
 'power-bowrider',
 'power-bass',
 'power-center',
 'power-cruiser']
In [32]:
#Checking make with log_price and condition
top_5(data,'make','log_price','condition')
Out[32]:
['Tracker', 'Sun Tracker', 'Bennington', 'Yamaha Boats', 'Sea Ray']
In [33]:
#checking hull material with price
top_5(data,'hullMaterial','log_price','condition')
Out[33]:
['fiberglass', 'other', 'aluminum', 'composite', 'wood']
In [38]:
#fuel type and price
sns.catplot(x="fuelType", y="log_price",data=data)
Out[38]:
<seaborn.axisgrid.FacetGrid at 0x24f10aab588>
In [39]:
#age and fuel_type with price
plt.figure(figsize=(10,5))
scatter = sns.scatterplot(data=data, x='age', y='log_price',hue ="fuelType")
In [87]:
#Statewise count
plt.figure(figsize=(20,50))
df_plot = data.groupby(['state', 'condition']).size().sort_values(ascending = False).reset_index().pivot(columns='condition', index='state', values=0)
df_plot["sum"] = df_plot['new'] + df_plot["used"]
df_plot = df_plot.sort_values(by=['sum'])
df_plot = df_plot.drop(columns=['sum'])

#plt.figure(figsize=(20,50))
df_plot.plot(kind='barh', stacked=True, figsize= (20,50))
plt.show()

<Figure size 1440x3600 with 0 Axes>
In [90]:
#HP and Price with new and old boats
plt.figure(figsize=(10,5))

scatter = sns.scatterplot(data=data, x='totalHP', y='log_price')
#Exploring more on engine category here
In [91]:
sns.catplot(x="engineCategory", y="log_price", hue = "".data=data)
Out[91]:
<seaborn.axisgrid.FacetGrid at 0x24f118237c8>
In [76]:
df_plot = data.groupby(['state', 'condition']).size().sort_values(ascending = False).reset_index()
In [92]:
data["make"].value_counts()
Out[92]:
Tracker               1894
Sun Tracker            907
Bennington             843
Yamaha Boats           728
Sea Ray                728
                      ... 
Chris Craft Launch       1
Motor Surf Boat          1
Northcoast Yachts        1
Encore                   1
Shallow Stalker          1
Name: make, Length: 960, dtype: int64