To explore the relationship between athlete attrbiutes and injury likelihood, we first examine age, weight, and height using data aggregation and visualization.

Analyzing by Age Groups

Method 1: Pivot Table

age_df = pd.pivot_table(df, values='Recovery_Time', index='Player_Age', columns='Likelihood_of_Injury', aggfunc='count')

# Rename columns using a dictionary
injury_dict = {0: 'NoLikelihood_Num', 1: 'Likelihood_Num'}
age_df.columns = age_df.columns.map(injury_dict)

# Calculate percentages
total = age_df['NoLikelihood_Num'] + age_df['Likelihood_Num']
age_df['NoLikelihood_Per'] = round(age_df['NoLikelihood_Num'] / total * 100, 2)
age_df['Likelihood_Per'] = round(age_df['Likelihood_Num'] / total * 100, 2)

Method 2: GroupBy with Apply

def age_per(x):
    total = x.shape[0]
    likelihood_num = (x['Likelihood_of_Injury'] == 1).sum()
    nolikelihood_num = (x['Likelihood_of_Injury'] == 0).sum()
    likelihood_per = round(likelihood_num / total * 100, 2)
    nolikelihood_per = round(nolikelihood_num / total * 100, 2)
    return pd.Series({'Likelihood_Per': likelihood_per, 'NoLikelihood_Per': nolikelihood_per})

age_df = age_df.groupby(by=['Age_Q'], as_index=False).apply(age_per)
display(age_df)

Visualization

sns.set_palette(my_palette)
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.figure(figsize=(8, 5))

ax = sns.barplot(x='Age_Q', y='Likelihood_Per', data=age_df, label='Likelihood', width=0.2)
sns.barplot(x='Age_Q', y='NoLikelihood_Per', data=age_df, label='NoLikelihood', ax=ax, width=0.06)

ax.set_title('Percentage of Athletes by Age Group and Injury Likelihood', pad=10, fontsize=14)
ax.set_xlabel('Age_Q')
ax.set_ylabel('per(%)')

for idx, row in age_df.iterrows():
    ax.text(row.name, row['Likelihood_Per'], str(row['Likelihood_Per']) + '%', ha='center', va='bottom', fontsize=10)
    ax.text(row.name, row['NoLikelihood_Per'], str(row['NoLikelihood_Per']) + '%', ha='center', va='bottom', fontsize=10)

plt.legend(loc=4)
plt.show()

Insight: Young adulthood is often considered a prime period in sports; training intensity may be higher, leading to a greater chance of injury.

Analyzing by Weight Groups

weight_df = df.copy()

# Method 1: Quantile-based bins (may lack practical meaning)
weight_bins = np.quantile(weight_df['Player_Weight'], np.linspace(0, 1, 5))
weight_df['Weight_Q'] = pd.cut(weight_df['Player_Weight'], bins=weight_bins, labels=['40-68kg', '68-75kg', '75-81kg', '81-105kg'])

# Method 2: Manual quartile boundaries
weight_max = weight_df['Player_Weight'].max()
weight_min = weight_df['Player_Weight'].min()
weight_q2 = weight_df['Player_Weight'].quantile(0.25)
weight_q3 = weight_df['Player_Weight'].quantile(0.5)
weight_q4 = weight_df['Player_Weight'].quantile(0.75)
weight_bins = [weight_min, weight_q2, weight_q3, weight_q4, weight_max]
weight_df['Weight_Q'] = pd.cut(weight_df['Player_Weight'], bins=weight_bins, labels=['40-68kg', '68-75kg', '75-81kg', '81-105kg'])

# Method 3: Custom bins
weight_df['Weight_Q'] = pd.cut(weight_df['Player_Weight'], bins=[0, 50, 75, 90, 110], labels=['<50kg', '50-75kg', '75-90kg', '>90kg'])

def weight_per(x):
    total = x.shape[0]
    likelihood_num = (x['Likelihood_of_Injury'] == 1).sum()
    nolikelihood_num = (x['Likelihood_of_Injury'] == 0).sum()
    likelihood_per = round(likelihood_num / total * 100, 2)
    nolikelihood_per = round(nolikelihood_num / total * 100, 2)
    return pd.Series({'Likelihood_Per': likelihood_per, 'NoLikelihood_Per': nolikelihood_per})

weight_df = weight_df.groupby(by='Weight_Q', as_index=False).apply(weight_per)
display(weight_df)

# Visualization
sns.set_palette(my_palette)
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.figure(figsize=(8, 5))

ax = sns.barplot(x='Weight_Q', y='Likelihood_Per', data=weight_df, width=0.2, label='Likelihood_Per')
sns.barplot(x='Weight_Q', y='NoLikelihood_Per', data=weight_df, ax=ax, width=0.06, label='NoLikelihood_Per')

for idx, row in weight_df.iterrows():
    plt.text(row.name, row['Likelihood_Per'], 'Green: ' + str(row['Likelihood_Per']) + '%', ha='center', va='bottom')
    plt.text(row.name, row['NoLikelihood_Per'], 'Red: ' + str(row['NoLikelihood_Per']) + '%', ha='center', va='top')

plt.title('Percentage of Athletes by Weight Group and Injury Likelihood', pad=10, fontsize=14)
plt.legend(loc=4)
plt.show()

Insight: Athletes under 50kg have an 83.3% chance of injury, indicating that lighter athletes are more prone to injuries.

Analyzing by Height Groups

height_df = df.copy()

# Quantile-based bins (may lack practical meaning)
height_bins = np.quantile(height_df['Player_Height'], np.linspace(0, 1, 5))
height_df['Height_Q'] = pd.cut(height_df['Player_Height'], bins=height_bins, labels=['145-173cm', '173-180cm', '180-187cm', '187-207cm'])

# Custom bins
height_df['Height_Q'] = pd.cut(height_df['Player_Height'], bins=[0, 165, 175, 185, 210], labels=['<165cm', '165-175cm', '175-185cm', '>185cm'])

def height_per(x):
    total = x.shape[0]
    likelihood_num = (x['Likelihood_of_Injury'] == 1).sum()
    nolikelihood_num = (x['Likelihood_of_Injury'] == 0).sum()
    likelihood_per = round(likelihood_num / total * 100, 2)
    nolikelihood_per = round(nolikelihood_num / total * 100, 2)
    return pd.Series({'Likelihood_Per': likelihood_per, 'NoLikelihood_Per': nolikelihood_per})

height_df = height_df.groupby(by='Height_Q', as_index=False).apply(height_per)
display(height_df)

# Visualization
sns.set_palette(my_palette)
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.figure(figsize=(8, 5))

ax = sns.barplot(x='Height_Q', y='Likelihood_Per', data=height_df, width=0.2, label='Likelihood_Per')
sns.barplot(x='Height_Q', y='NoLikelihood_Per', data=height_df, ax=ax, width=0.06, label='NoLikelihood_Per')

for idx, row in height_df.iterrows():
    plt.text(row.name, row['Likelihood_Per'], str(row['Likelihood_Per']) + '%', ha='center', va='bottom', fontsize=10)
    plt.text(row.name, row['NoLikelihood_Per'], str(row['NoLikelihood_Per']) + '%', ha='center', va='bottom', fontsize=10)

plt.title('Percentage of Athletes by Height Group and Injury Likelihood', pad=10, fontsize=14)
plt.legend(loc=4)
plt.show()

Insight: Athletes in the 175-185cm range appear more prone to injury compared to other height groups.

Step 2: Feature Correlation Analysis

We use Pearson correlation coefficient.

correlation_matrix = df.corr(method='pearson')
display(correlation_matrix)

plt.rcParams['font.sans-serif'] = ['SimHei']
plt.figure(figsize=(6, 5))

sns.heatmap(correlation_matrix, annot=True, cmap='coolwarm', linewidth=0.5)
plt.title('Correlation Heatmap of Athlete Attributes and Injury Likelihood', pad=20)
plt.show()

Summary: The correlation between injury likelihood and training intensity is 0.089, indicating a weak positive relationship.

Step 3: Feature Importance Analysis

We use a Random Forest classifier to evaluate feature importance.

# Prepare data
X, y = df.iloc[:, :-1], df['Likelihood_of_Injury']

# Split into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Train model
model = RandomForestClassifier()
model.fit(X_train, y_train)

# Predict and evaluate
predictions = model.predict(X_test)
print(f"Model prediction accuracy: {round(accuracy_score(y_test, predictions), 2) * 100}%\n")

# Retrain on full dataset to get feature importances
model.fit(X, y)
feature_importances = (model.feature_importances_ * 100).round(2)

fi_df = pd.DataFrame(data=feature_importances, index=df.columns[:-1], columns=['Importance']).sort_values('Importance', ascending=False)

# Visualization
sns.set_palette(my_palette)
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.figure(figsize=(8, 5))
plt.title('Random Forest Feature Importances', pad=20)

sns.barplot(x='Importance', y=fi_df.index, data=fi_df, width=0.5)

for idx, row in fi_df.iterrows():
    plt.text(row['Importance'], row.name, str(row['Importance']) + '%', ha='right', va='center')

plt.show()

The model achieves 53% accuracy, suggesting other models might be more suitable. The feature importance plot shows that training intensity, weight, and height are relatively important, which aligns with earlier analyses.