利用决策树预测NBA比赛胜负的Python数据挖掘实践

### 1. 数据准备与预处理

本文的数据来源包括2013-14赛季NBA的比赛记录和2013年的球队排名。我们首先加载并初步检查这些数据。

```python
import pandas as pd
import time

start = time.time()
file_path = 'path_to_your_data/NBA_2014_games.csv'
data = pd.read_csv(file_path)
print(data.head())
print(f'Data shape: {data.shape}')
print('Duplicate rows:', data[data.duplicated()])
```

接下来，对数据进行清洗，确保日期格式正确，并重命名列标题以提高可读性。

```python
data = pd.read_csv(file_path, parse_dates=['Date'])
data.columns = ['Date', 'Score Type', 'Visitor Team', 'VisitorPts', 'Home Team', 'HomePts', 'OT?', 'Notes']
print(data.head())
```

### 2. 特征工程

为了更好地训练模型，我们需要从原始数据中提取有用的信息作为新特征。例如，我们可以根据历史战绩判断主场或客场比赛的结果。

```python
data['HomeWin'] = data['VisitorPts'] y_true = data['HomeWin'].values

from collections import defaultdict
won_last = defaultdict(int)
data['HomeLastWin'], data['VisitorLastWin'] = None, None

for index, row in data.iterrows():
home_team, visitor_team = row['Home Team'], row['Visitor Team']
data.at[index, 'HomeLastWin'] = won_last[home_team]
data.at[index, 'VisitorLastWin'] = won_last[visitor_team]
won_last[home_team] = row['HomeWin']
won_last[visitor_team] = not row['HomeWin']

print(data[['Home Team', 'HomePts', 'HomeWin', 'HomeLastWin', 'VisitorLastWin']].head())
```

### 3. 模型构建与评估

这里选择决策树作为基础分类器，并通过交叉验证评估其性能。

```python
from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import cross_val_score
import numpy as np

clf = DecisionTreeClassifier(random_state=14)
X_previousWins = data[['HomeLastWin', 'VisitorLastWin']].values
scores = cross_val_score(clf, X_previousWins, y_true, scoring='accuracy')
mean_accuracy = np.mean(scores) * 100
print(f'Accuracy using previous wins: {mean_accuracy:.2f}%')
```

此外，还可以引入更多特征（如球队排名差异）来改进模型表现。

```python
standings = pd.read_csv('path_to_your_data/NBA_2013_standings.csv')
data['HomeTeamRanksHigher'] = 0

for index, row in data.iterrows():
home_team, visitor_team = row['Home Team'], row['Visitor Team']
if home_team == 'New Orleans Pelicans':
home_team = 'New Orleans Hornets'
elif visitor_team == 'New Orleans Pelicans':
visitor_team = 'New Orleans Hornets'
home_rank = standings[standings['Team'] == home_team]['Rk'].values[0]
visitor_rank = standings[standings['Team'] == visitor_team]['Rk'].values[0]
data.at[index, 'HomeTeamRanksHigher'] = int(home_rank > visitor_rank)

X_homehigher = data[['HomeLastWin', 'VisitorLastWin', 'HomeTeamRanksHigher']].values
scores = cross_val_score(clf, X_homehigher, y_true, scoring='accuracy')
mean_accuracy_with_rank = np.mean(scores) * 100
print(f'Improved accuracy with ranking: {mean_accuracy_with_rank:.2f}%')
```

最后，尝试使用随机森林进一步优化预测效果，并通过网格搜索确定最佳参数组合。

```python
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import GridSearchCV

param_grid = {
'max_features': [2, 3, 'auto'],
'n_estimators': [100, 110, 120],
'criterion': ['gini', 'entropy'],
'min_samples_leaf': [2, 4, 6]
}

rf_clf = RandomForestClassifier(random_state=14, n_jobs=-1)
grid_search = GridSearchCV(rf_clf, param_grid)
grid_search.fit(X_homehigher, y_true)
optimal_accuracy = grid_search.best_score_ * 100
print(f'Optimized accuracy: {optimal_accuracy:.2f}%')
print('Best parameters:', grid_search.best_params_)

end = time.time()
print(f'Total runtime: {end - start:.2f} seconds')
```

以上代码展示了如何利用Python及其相关库完成一个完整的数据挖掘项目流程，包括但不限于数据获取、清理、特征构建、模型训练及调优等方面的工作。