added risk analysis stuff

alphamind/analysis/__init__.py

+# -*- coding: utf-8 -*-
+"""
+Created on 2017-5-6
+
+@author: cheng.li
+"""

alphamind/analysis/riskanalysis.py

+# -*- coding: utf-8 -*-
+"""
+Created on 2017-5-6
+
+@author: cheng.li
+"""
+
+from typing import Tuple
+import numpy as np
+import pandas as pd
+from alphamind.data.neutralize import neutralize
+
+
+def risk_analysis(return_series: pd.Series,
+                  risk_return_table: pd.DataFrame) -> Tuple[pd.DataFrame, pd.DataFrame]:
+
+    group_idx = return_series.index.values.astype(int)
+    return_values = return_series.values
+    risk_return_values = risk_return_table.values
+    risk_factor_cols = risk_return_table.columns
+
+    idiosyncratic, other_stats = neutralize(risk_return_values,
+                                            return_values,
+                                            group_idx,
+                                            output_exposure=True,
+                                            output_explained=True)
+
+    systemetic = other_stats['explained']
+    exposure = other_stats['exposure']
+
+    explained_table = np.hstack((idiosyncratic, systemetic[:, :, 0]))
+    cols = ['idiosyncratic']
+    cols.extend(risk_factor_cols)
+
+    explained_table = pd.DataFrame(explained_table, columns=cols, index=return_series.index)
+    exposure_table = pd.DataFrame(exposure[:, :, 0], columns=risk_factor_cols, index=return_series.index)
+    return explained_table, exposure_table.groupby(level=0).first()
+
+
+if __name__ == '__main__':
+    from matplotlib import pyplot as plt
+
+    n_samples = 360000
+    n_dates = 200
+    n_risk_factors = 35
+
+    dates = np.sort(np.random.randint(n_dates, size=n_samples))
+    return_series = pd.Series(data=np.random.randn(n_samples), index=dates)
+    risk_return_table = pd.DataFrame(data=np.random.randn(n_samples, n_risk_factors),
+                                     columns=list(range(n_risk_factors)),
+                                     index=dates)
+
+    explained_table, exposure_table = risk_analysis(return_series, risk_return_table)
+
+    aggregated_bars = explained_table.groupby(level=0).sum()
+    top_sources = aggregated_bars.sum().abs().sort_values(ascending=False).index[:10]
+    aggregated_bars.sum().sort_values(ascending=False).plot(kind='bar', figsize=(16, 8))
+
+    exposure_table[top_sources.difference(['idiosyncratic'])].plot(figsize=(14, 7))
+
+    plt.show()
+