correlation-explorer

相関エクスプローラー

CSV／Excel データセット内の変数間の相関を分析します。

機能

• 相関行列：すべてのペア間の相関を計算します
• ヒートマップによる可視化：色分けして相関を表示します
• 有意性検定：相関の p 値を算出します
• 複数の手法：Pearson、Spearman、Kendall
• 強い相関：相関の強いペアを検出します
• ターゲット分析：特定の変数との相関を分析します

クイックスタート

from correlation_explorer import CorrelationExplorer

explorer = CorrelationExplorer()

# Load and analyze
explorer.load_csv("sales_data.csv")
matrix = explorer.correlation_matrix()

# Find strong correlations
strong = explorer.find_strong_correlations(threshold=0.7)
print(strong)

# Generate heatmap
explorer.plot_heatmap("correlation_heatmap.png")

CLI の使い方

# Compute correlation matrix
python correlation_explorer.py --input data.csv --output correlations.csv

# Generate heatmap
python correlation_explorer.py --input data.csv --heatmap heatmap.png

# Find strong correlations
python correlation_explorer.py --input data.csv --strong --threshold 0.7

# Correlations with target variable
python correlation_explorer.py --input data.csv --target sales

# Use Spearman correlation
python correlation_explorer.py --input data.csv --method spearman

# Include p-values
python correlation_explorer.py --input data.csv --pvalues

API リファレンス

CorrelationExplorer クラス

class CorrelationExplorer:
    def __init__(self)

    # Data loading
    def load_csv(self, filepath: str, **kwargs) -> 'CorrelationExplorer'
    def load_dataframe(self, df: pd.DataFrame) -> 'CorrelationExplorer'

    # Analysis
    def correlation_matrix(self, method: str = "pearson") -> pd.DataFrame
    def correlation_with_pvalues(self, method: str = "pearson") -> tuple
    def correlate_with_target(self, target: str, method: str = "pearson") -> pd.Series

    # Discovery
    def find_strong_correlations(self, threshold: float = 0.7) -> list
    def find_weak_correlations(self, threshold: float = 0.3) -> list

    # Visualization
    def plot_heatmap(self, output: str, **kwargs) -> str
    def plot_scatter(self, var1: str, var2: str, output: str) -> str

    # Export
    def to_csv(self, output: str) -> str
    def to_json(self, output: str) -> str

相関手法

手法	適した用途
pearson	線形関係、正規分布データ
spearman	非線形関係、順序データ
kendall	小規模サンプル、順序データ

# Pearson (default) - parametric
matrix = explorer.correlation_matrix(method="pearson")

# Spearman - rank-based, non-parametric
matrix = explorer.correlation_matrix(method="spearman")

# Kendall - robust to outliers
matrix = explorer.correlation_matrix(method="kendall")

出力形式

相関行列

           sales  marketing  customers
sales      1.000      0.854      0.723
marketing  0.854      1.000      0.612
customers  0.723      0.612      1.000

強い相関

[
    {"var1": "sales", "var2": "marketing", "correlation": 0.854, "abs_corr": 0.854},
    {"var1": "sales", "var2": "customers", "correlation": 0.723, "abs_corr": 0.723}
]

p 値付き

{
    "correlations": DataFrame,
    "pvalues": DataFrame,
    "significant": [...],  # p < 0.05
}

ワークフロー例

特徴量選択

explorer = CorrelationExplorer()
explorer.load_csv("features.csv")

# Find features correlated with target
target_corr = explorer.correlate_with_target("target")
important_features = target_corr[abs(target_corr) > 0.3].index.tolist()
print(f"Important features: {important_features}")

# Find multicollinear features (to potentially drop)
strong = explorer.find_strong_correlations(threshold=0.9)
print("Highly correlated pairs (consider dropping one):")
for pair in strong:
    print(f"  {pair['var1']} <-> {pair['var2']}: {pair['correlation']:.3f}")

売上分析

explorer = CorrelationExplorer()
explorer.load_csv("sales_data.csv")

# What drives sales?
sales_corr = explorer.correlate_with_target("revenue")
print("Factors correlated with revenue:")
for var, corr in sales_corr.sort_values(ascending=False).items():
    if var != "revenue":
        print(f"  {var}: {corr:.3f}")

# Visualize
explorer.plot_heatmap("sales_correlations.png")

データ探索

explorer = CorrelationExplorer()
explorer.load_csv("dataset.csv")

# Get full picture
corr, pvals = explorer.correlation_with_pvalues()

# Find all significant correlations
significant = []
for i in range(len(corr.columns)):
    for j in range(i+1, len(corr.columns)):
        if pvals.iloc[i, j] < 0.05:
            significant.append({
                'var1': corr.columns[i],
                'var2': corr.columns[j],
                'r': corr.iloc[i, j],
                'p': pvals.iloc[i, j]
            })

ヒートマップのオプション

explorer.plot_heatmap(
    output="heatmap.png",
    cmap="coolwarm",      # Color scheme
    annot=True,           # Show values
    figsize=(12, 10),     # Figure size
    vmin=-1, vmax=1,      # Color scale
    title="Correlation Matrix"
)

依存関係

• pandas>=2.0.0
• numpy>=1.24.0
• scipy>=1.10.0
• matplotlib>=3.7.0
• seaborn>=0.12.0