
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sb
import folium
from sklearn import linear_model
import statsmodels.api as sm
import statsmodels.formula.api as smf
from statistics import mean
from sklearn import tree
from sklearn.model_selection import train_test_split
from sklearn import metrics


olympics_df = pd.read_csv('athlete_events.csv')
olympics_df.head()


# create top_df to contain data consisting of the same time period, but only 11 countries and NaN values dropped

# copy the original olympics_df to a new one called top_df (representing data of the top medal winning countries)
top_df = olympics_df.copy()

# drop all the rows with NaN values in the medal, height, or weight columns
top_df = top_df[top_df['Medal'].notna()]
top_df = top_df[top_df['Height'].notna()]
top_df = top_df[top_df['Weight'].notna()]

# create a list of the abbreviated names of countries we will be focusing on
countries = ['USA', 'RUS', 'GBR', 'FRA', 'CHN', 'ITA', 'AUS', 'SWE', 'HUN', 'JPN', 'GER']

# iterate through top_df and drop rows pertaining to countries that are not in the countries list
for index,row in top_df.iterrows():
    if row['NOC'] not in countries:
        top_df = top_df.drop(index, 0)

# reset index values after removing rows
top_df.reset_index(drop=True, inplace=True)

# display data
top_df


# time_pds_df contains data of the first five and last five modern Olympics
time_pds_df = top_df.copy()

# create list of the years in focus
years = [1896, 1900, 1904, 1908, 1912, 2000, 2004, 2008, 2012, 2016]

# tidy further by dropping any rows that are not Summer Olympics
time_pds_df = time_pds_df[time_pds_df['Season'] == 'Summer'].reset_index(drop=True)

# iterate through time_pds_df and drop rows containing data not within the time periods in focus
for index,row in time_pds_df.iterrows():
    if row['Year'] not in years:
        time_pds_df = time_pds_df.drop(index, 0)

# reset index after dropping rows in df
time_pds_df.reset_index(drop=True, inplace=True)

# display data
time_pds_df


# create world map pinpointing the eleven highest medal earning countries
map_osm = folium.Map()

# create df with country, country's capital name, and the capital's longitude and latitude values using values from map_data
map_data = [['USA', 'Washington DC, USA', 38.8899, -77.0091], ['RUS','Moscow, Russia', 55.7558, 37.6173], ['GBR','London, UK', 51.5072, 0.1276], ['FRA','Paris, France',48.8566, 2.3522], ['CHN', 'Beijing, China', 39.9042, 116.4074], ['ITA', 'Rome, Italy',41.9028, 12.4964], ['AUS', 'Canberra, Australia', -35.2809, 149.1300], ['SWE','Stockholm, Sweden', 59.3293, 18.0686],
       ['HUN', 'Budapest, Hungary', 47.4979, 19.0402], ['JPN', 'Tokyo, Japan', 35.6762, 139.6503], ['GER', 'Berlin, Germany', 52.5200, 13.4050]]

# columns of the new df are NOC, capital, latitude, and longitude
loc_df = pd.DataFrame(map_data, columns = ['NOC', 'Capital', 'Latitude', 'Longitude'])

# create new column that currently stores NaN values, but will be updated to store each country's total medal count
loc_df['Medal_Count'] = np.nan

# iterate through top_df and sum up each country's total number of medals, storing them in medal_count_arr
# medal_count_arr is a list for storing each country's medal count, where each country is assigned to a numerical index/key
medal_count_arr = [0] * 11

for index, row in top_df.iterrows():
    if row['NOC'] == 'USA':
        medal_count_arr[0] += 1
    elif row['NOC'] == 'RUS':
        medal_count_arr[1] += 1
    elif row['NOC'] == 'GBR':
        medal_count_arr[2] += 1
    elif row['NOC'] == 'FRA':
        medal_count_arr[3] += 1
    elif row['NOC'] == 'CHN':
        medal_count_arr[4] += 1
    elif row['NOC'] == 'ITA':
        medal_count_arr[5] += 1
    elif row['NOC'] == 'AUS':
        medal_count_arr[6] += 1
    elif row['NOC'] == 'SWE':
        medal_count_arr[7] += 1
    elif row['NOC'] == 'HUN':
        medal_count_arr[8] += 1
    elif row['NOC'] == 'JPN':
        medal_count_arr[9] += 1
    elif row['NOC'] == 'GER':
        medal_count_arr[10] += 1
        
# update loc_df medal count column with the medal_count_arr values corresponding to each country
for index, row in loc_df.iterrows():
    loc_df.at[index,'Medal_Count'] = medal_count_arr[index]

# display loc_df
loc_df


# add each country's marker to the map with color associated with continent & include a pop-up containing location name and medal count
for indice, row in loc_df.iterrows():
    # assign country an icon color
    if row['NOC'] == 'USA': # north america
        color = 'red'
    elif row['NOC'] == 'RUS' or row['NOC'] == 'GBR' or row['NOC'] == 'FRA' or row['NOC'] == 'ITA' or row['NOC'] == 'HUN' or row['NOC'] == 'GER' or row['NOC'] == 'SWE': # europe
        color = 'blue'
    elif row['NOC'] == 'CHN' or row['NOC'] == 'JPN': # asia
        color = 'green'
    elif row['NOC'] == 'AUS': # austrailia
        color = 'orange'
        
    # add marker to map 
    folium.Marker(
        location=[row["Latitude"], row["Longitude"]],
        popup= str(row["Capital"]) + '\nMedal Count: ' + str(row["Medal_Count"]),
        icon=folium.map.Icon(color=color)
    ).add_to(map_osm)
    
# display map
map_osm


# list of variable names to measure 
variables = ["Height", "Weight", "Age"]

# displays graphs of the target variable vs time
fig3, axes = plt.subplots(len(variables), 1, figsize=(15,20), squeeze=False)
plt.subplots_adjust(hspace=.4)

# creates a scatterplot using matplotlib pyplot for each variable vs year
for i in range(len(variables)):
    top_df.plot(kind='scatter', x="Year", y=variables[i], ax=axes[i, 0],
                    title = variables[i] + " vs Year")

# display graphs
plt.show()


# create new dataframes, that are copies of top_df, to be manipulated
early_df = top_df.copy()
later_df = top_df.copy()

# early_df consists of data from before 1950, later_df consists of data from 1950 and later
# reset index after dropping rows from both df
early_df = early_df[early_df['Year'] < 1950].reset_index(drop=True)
later_df = later_df[later_df['Year'] >= 1950].reset_index(drop=True)

# create pie chart using matplotlib pyplot that plots number of males and females from the earlier Olympic years (before 1950)
plt.figure(figsize=[9,7])
early_df['Sex'].value_counts().plot.pie()
plt.title('Male and Female Medalists Count from 1896 to 1950')

# display pie chart
plt.show()


# create pie chart using matplotlib pyplot that plots number of males and females from the later Olympic years (1950 and later)
plt.figure(figsize=[9,7])
later_df['Sex'].value_counts().plot.pie()
plt.title('Male and Female Medalists Count from 1950 to 2016')

# display pie chart
plt.show()


# create a dictionary that will store key/value pairs for female count for a given year
fem_yr = dict()

# create copy of top_df stored in female_df, then keep only data pertaining to female athletes
female_df = top_df.copy()
female_df = female_df[female_df['Sex'] == 'F'].reset_index(drop=True)

# iterate through female_df to count up the number of females in a given year
for index, row in female_df.iterrows():       
    if row['Year'] in fem_yr:
        fem_yr[row['Year']] = fem_yr[row['Year']] + 1
    else:
        fem_yr[row['Year']] = 1
        
# create lists for the keys and values of the fem_yr dict
year_lst = []
female_count = []
for k,v in fem_yr.items():
    year_lst.append(k)
    female_count.append(v)

# create a scatterplot using matplotlib pyplot of female medalists count over time
plt.scatter(year_lst, female_count, s=50)
plt.title("Number of Female Medalists Over Time", fontsize = 12)
plt.xlabel('Year')
plt.ylabel('Count')

# add regression line
m, b = np.polyfit(year_lst, female_count, 1)
y = female_df['Year']
plt.plot(y, m*y+b, color='red')

# print slope of line that shows numerical rate of increase
print('Slope of line: ' + str(m))

Slope of line: 4.717016778966828


# create a dictionary that will store key/value pairs for male count for a given year
male_yr = dict()

# create copy of top_df stored in male_df, then keep only data pertaining to male athletes
male_df = top_df.copy()
male_df = male_df[male_df['Sex'] == 'M'].reset_index(drop=True)

# iterate through male_df to count up the number of males in a given year
for index, row in male_df.iterrows():       
    if row['Year'] in male_yr:
        male_yr[row['Year']] = male_yr[row['Year']] + 1
    else:
        male_yr[row['Year']] = 1
        
# create lists for the keys and values of the male_yr dict
year_lst2 = []
male_count = []
for k,v in male_yr.items():
    year_lst2.append(k)
    male_count.append(v)
    
# create a scatterplot using matplotlib pyplot of male medalists count over time
plt.scatter(year_lst2, male_count, s=50)
plt.title("Number of Male Medalists Over Time", fontsize = 12)
plt.xlabel('Year')
plt.ylabel('Count')

# add regression line
m2, b2 = np.polyfit(year_lst2, male_count, 1)
y2 = male_df['Year']
plt.plot(y2, m2*y2+b2, color='red')

# print slope of line that shows numerical rate of increase
print('Slope of line: ' + str(m2))

Slope of line: 3.3164683242687882


# create a boxplot using seaborn for plotting categorical sex variable over time (years)
sb.boxplot(x='Year', y='Sex', data=top_df, width = 0.8).set_title('Sex vs. Year of Olympic Athletes from 1986 to 2016')

Text(0.5, 1.0, 'Sex vs. Year of Olympic Athletes from 1986 to 2016')


# copy top_df and store in new reg_df
reg_df = top_df.copy()

# convert 'Sex' column's string values (M/F) to numerical values (0/1) to be used in regression model
reg_df.replace(to_replace= 'M', value = 0, inplace=True)
reg_df.replace(to_replace= 'F', value = 1, inplace=True)

# create arrays of year and sex values needed to create the regression model
xplot = np.array(reg_df['Year'].values)
yplot = np.array(reg_df['Sex'].values)

# fit a linear regression model using statsmodels for year and sex
model = sm.OLS(yplot, xplot, missing='drop')
model = model.fit()

# print out ols regression result summary for inspection, including model coefficient and intercept
model.summary()


# create plot using matplotlib pyplot
fig, ax = plt.subplots()
fig.set_figheight(6)
fig.set_figwidth(6)

# iterate through the dataframe and add the key value pairs of year and age to the age_yr dictionary
age_yr = dict()

for index, row in time_pds_df.iterrows():
    if row['Year'] < 2000:
        if row['Year'] in age_yr:
            age_yr[row['Year']].append(row['Age'])
        else:
            age_yr[row['Year']] = [row['Age']]

# create lists for the keys and values of the age_yr dict
x = []
y = []
for k,v in age_yr.items():
    x.append(k)
    y.append(v)

# create violin plot for x (year) and y (age)
ax.violinplot(y,x,widths=2,showmeans=True)
ax.set_xlabel("Year")
ax.set_ylabel("Age")
ax.set_title("Distribution of Age Over Time: 1896-1910")
fig.savefig("violin.png")


# create plot using matplotlib pyplot
fig2, ax2 = plt.subplots()
fig2.set_figheight(6)
fig2.set_figwidth(6)

# iterate through the dataframe and add the key value pairs of year and age to the age_yr dictionary
age_yr2 = dict()
for index, row in time_pds_df.iterrows():
    if row['Year'] >= 2000:
        if row['Year'] in age_yr2:
            age_yr2[row['Year']].append(row['Age'])
        else:
            age_yr2[row['Year']] = [row['Age']]

# create lists for the keys and values of the age_yr2 dict
x2 = []
y2 = []
for k,v in age_yr2.items():
    x2.append(k)
    y2.append(v)

# create violin plot for x (year) and y (age)
ax2.violinplot(y2,x2,widths=2,showmeans=True)
ax2.set_xlabel("Year")
ax2.set_ylabel("Age")
ax2.set_title("Distribution of Age Over Time: 2000-2016")
fig2.savefig("violin.png")


# create a distplot using seaborn which represents the distribution of age against the density distribution
sb.distplot(time_pds_df.Age)

/opt/conda/lib/python3.8/site-packages/seaborn/distributions.py:2557: FutureWarning: `distplot` is a deprecated function and will be removed in a future version. Please adapt your code to use either `displot` (a figure-level function with similar flexibility) or `histplot` (an axes-level function for histograms).
  warnings.warn(msg, FutureWarning)

<AxesSubplot:xlabel='Age', ylabel='Density'>


# create new column in reg_df for age group
reg_df['Age_Group'] = 'empty'

# iterate through reg_df to assign the athlete to an age group based on their age
# Teen: under 18, Youth: 18-24, Adult: 25-54, Senior: 55 and up
for index, row in reg_df.iterrows():
    if row['Age'] < 18:
        reg_df.at[index, 'Age_Group'] = 'Teen'
    elif row['Age'] >= 18 and row['Age'] < 25:
        reg_df.at[index, 'Age_Group'] = 'Youth'
    elif row['Age'] >= 25 and row['Age'] < 55:
        reg_df.at[index, 'Age_Group'] = 'Adult'
    else:
        reg_df.at[index, 'Age_Group'] = 'Senior'  

# create bar graph showing athlete count for each age group
plt.figure(figsize=[9,7])
reg_df['Age_Group'].value_counts().plot.barh()
plt.title('Age Group vs. Athlete Count')
plt.xlabel('Count')
plt.ylabel('Age Group')

# display plot
plt.show()


# remove any NaN values that may be in the data
reg_df = reg_df[~reg_df.isin([np.nan, np.inf, -np.inf]).any(1)]

# decision tree is the algorithm used for classification
# using entropy (range of 0 to 1) as the criterion for measuring the quality of the split and selecting the best features
# for the max_depth hyperparameter, we chose the value 5 versus default value of None which would allow nodes to expand
# until all leaves are pure or until all leaves contain less than min_samples_split samples; after testing different 
# values, max_depth of 3 consistently produced higher accuracy values.
dtc = tree.DecisionTreeClassifier(criterion = 'entropy', max_depth=3)

# the predictors to be used include age, height, and weight
# store the data and the target data into two variables
cols = ['Age', 'Height', 'Weight']
X = reg_df[cols] # input features
y = reg_df.Sex # target variable

# perform holdout validation on the algorithm
# split data for separate training and testing sets, where 70% of dataset is used for training, and 30% for testing
X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=.3)

# fit the model, train the decision tree classifier
dtc = dtc.fit(X_train,y_train)

# predict the test dataset using predict()
y_pred = dtc.predict(X_test)

# the performance metric used is accuracy (using metrics from sklearn)
dtc_accuracy = metrics.accuracy_score(y_test, y_pred)

# print accuracy of decision tree classifier
print("Accuracy:", dtc_accuracy)

Accuracy: 0.7695945945945946

	ID	Name	Sex	Age	Height	Weight	Team	NOC	Games	Year	Season	City	Sport	Event	Medal
0	1	A Dijiang	M	24.0	180.0	80.0	China	CHN	1992 Summer	1992	Summer	Barcelona	Basketball	Basketball Men's Basketball	NaN
1	2	A Lamusi	M	23.0	170.0	60.0	China	CHN	2012 Summer	2012	Summer	London	Judo	Judo Men's Extra-Lightweight	NaN
2	3	Gunnar Nielsen Aaby	M	24.0	NaN	NaN	Denmark	DEN	1920 Summer	1920	Summer	Antwerpen	Football	Football Men's Football	NaN
3	4	Edgar Lindenau Aabye	M	34.0	NaN	NaN	Denmark/Sweden	DEN	1900 Summer	1900	Summer	Paris	Tug-Of-War	Tug-Of-War Men's Tug-Of-War	Gold
4	5	Christine Jacoba Aaftink	F	21.0	185.0	82.0	Netherlands	NED	1988 Winter	1988	Winter	Calgary	Speed Skating	Speed Skating Women's 500 metres	NaN

	ID	Name	Sex	Age	Height	Weight	Team	NOC	Games	Year	Season	City	Sport	Event	Medal
0	62	Giovanni Abagnale	M	21.0	198.0	90.0	Italy	ITA	2016 Summer	2016	Summer	Rio de Janeiro	Rowing	Rowing Men's Coxless Pairs	Bronze
1	67	Mariya Vasilyevna Abakumova (-Tarabina)	F	22.0	179.0	80.0	Russia	RUS	2008 Summer	2008	Summer	Beijing	Athletics	Athletics Women's Javelin Throw	Silver
2	73	Luc Abalo	M	23.0	182.0	86.0	France	FRA	2008 Summer	2008	Summer	Beijing	Handball	Handball Men's Handball	Gold
3	73	Luc Abalo	M	27.0	182.0	86.0	France	FRA	2012 Summer	2012	Summer	London	Handball	Handball Men's Handball	Gold
4	73	Luc Abalo	M	31.0	182.0	86.0	France	FRA	2016 Summer	2016	Summer	Rio de Janeiro	Handball	Handball Men's Handball	Silver
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
14800	135508	Vera Igorevna Zvonaryova	F	23.0	172.0	59.0	Russia	RUS	2008 Summer	2008	Summer	Beijing	Tennis	Tennis Women's Singles	Bronze
14801	135520	Julia Zwehl	F	28.0	167.0	60.0	Germany	GER	2004 Summer	2004	Summer	Athina	Hockey	Hockey Women's Hockey	Gold
14802	135525	Martin Zwicker	M	29.0	175.0	64.0	Germany	GER	2016 Summer	2016	Summer	Rio de Janeiro	Hockey	Hockey Men's Hockey	Bronze
14803	135563	Olesya Nikolayevna Zykina	F	19.0	171.0	64.0	Russia	RUS	2000 Summer	2000	Summer	Sydney	Athletics	Athletics Women's 4 x 400 metres Relay	Bronze
14804	135563	Olesya Nikolayevna Zykina	F	23.0	171.0	64.0	Russia	RUS	2004 Summer	2004	Summer	Athina	Athletics	Athletics Women's 4 x 400 metres Relay	Silver

	ID	Name	Sex	Age	Height	Weight	Team	NOC	Games	Year	Season	City	Sport	Event	Medal
0	62	Giovanni Abagnale	M	21.0	198.0	90.0	Italy	ITA	2016 Summer	2016	Summer	Rio de Janeiro	Rowing	Rowing Men's Coxless Pairs	Bronze
1	67	Mariya Vasilyevna Abakumova (-Tarabina)	F	22.0	179.0	80.0	Russia	RUS	2008 Summer	2008	Summer	Beijing	Athletics	Athletics Women's Javelin Throw	Silver
2	73	Luc Abalo	M	23.0	182.0	86.0	France	FRA	2008 Summer	2008	Summer	Beijing	Handball	Handball Men's Handball	Gold
3	73	Luc Abalo	M	27.0	182.0	86.0	France	FRA	2012 Summer	2012	Summer	London	Handball	Handball Men's Handball	Gold
4	73	Luc Abalo	M	31.0	182.0	86.0	France	FRA	2016 Summer	2016	Summer	Rio de Janeiro	Handball	Handball Men's Handball	Silver
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
6102	135508	Vera Igorevna Zvonaryova	F	23.0	172.0	59.0	Russia	RUS	2008 Summer	2008	Summer	Beijing	Tennis	Tennis Women's Singles	Bronze
6103	135520	Julia Zwehl	F	28.0	167.0	60.0	Germany	GER	2004 Summer	2004	Summer	Athina	Hockey	Hockey Women's Hockey	Gold
6104	135525	Martin Zwicker	M	29.0	175.0	64.0	Germany	GER	2016 Summer	2016	Summer	Rio de Janeiro	Hockey	Hockey Men's Hockey	Bronze
6105	135563	Olesya Nikolayevna Zykina	F	19.0	171.0	64.0	Russia	RUS	2000 Summer	2000	Summer	Sydney	Athletics	Athletics Women's 4 x 400 metres Relay	Bronze
6106	135563	Olesya Nikolayevna Zykina	F	23.0	171.0	64.0	Russia	RUS	2004 Summer	2004	Summer	Athina	Athletics	Athletics Women's 4 x 400 metres Relay	Silver

Dep. Variable:	y	R-squared (uncentered):	0.367
Model:	OLS	Adj. R-squared (uncentered):	0.367
Method:	Least Squares	F-statistic:	8580.
Date:	Sun, 19 Dec 2021	Prob (F-statistic):	0.00
Time:	07:26:45	Log-Likelihood:	-10130.
No. Observations:	14805	AIC:	2.026e+04
Df Residuals:	14804	BIC:	2.027e+04
Df Model:	1
Covariance Type:	nonrobust

	coef	std err	t	P>\|t\|	[0.025	0.975]
x1	0.0002	1.98e-06	92.627	0.000	0.000	0.000

From Ancient Greece to Your Screen: How the Olympics Have Developed Over Time¶

Emily Chang and Vidushi Vashist¶

Introduction¶

Getting Started with the Data¶

Data Collection and Wrangling¶

Exploratory Data Analysis¶

Map Overview of Countries of Focus¶

Height, Weight, and Age Distribution Over Time¶

Male and Female Olympic Athletes Over Time¶

Regression Model for Our Data¶

Distribution of Ages In Beginning Olympic Years vs Latter Olympic Years¶

Machine Learning¶

Conclusion¶

	NOC	Capital	Latitude	Longitude	Medal_Count
0	USA	Washington DC, USA	38.8899	-77.0091	4385.0
1	RUS	Moscow, Russia	55.7558	37.6173	1134.0
2	GBR	London, UK	51.5072	0.1276	1035.0
3	FRA	Paris, France	48.8566	2.3522	987.0
4	CHN	Beijing, China	39.9042	116.4074	985.0
5	ITA	Rome, Italy	41.9028	12.4964	1061.0
6	AUS	Canberra, Australia	-35.2809	149.1300	1207.0
7	SWE	Stockholm, Sweden	59.3293	18.0686	765.0
8	HUN	Budapest, Hungary	47.4979	19.0402	791.0
9	JPN	Tokyo, Japan	35.6762	139.6503	843.0
10	GER	Berlin, Germany	52.5200	13.4050	1612.0

Omnibus:	66488.601	Durbin-Watson:	1.225
Prob(Omnibus):	0.000	Jarque-Bera (JB):	2530.877
Skew:	0.568	Prob(JB):	0.00
Kurtosis:	1.323	Cond. No.	1.00