GroupBy: covert to DataFrame, OLS Regression, group predicti

# -*- coding: utf-8 -*-
"""
GroupBy: covert to DataFrame, OLS Regression, group predictions
Created on Mon Jun 12 20:57:32 2017
"""
import pandas as pd
import numpy as np
import statsmodels.api as sm
from sklearn.linear_model import LinearRegression

#%% Converting a Pandas GroupBy object to DataFrame
#https://stackoverflow.com/questions/10373660/
df1 = pd.DataFrame( {
  "Name" : ["Alice", "Bob", "Mallory", "Mallory", "Bob" , "Mallory"] ,
  "City" : ["Seattle", "Seattle", "Portland", "Seattle", "Seattle", "Portland"] } )

# g1 here is a DataFrame. It has a hierarchical index
g1 = df1.groupby( [ "Name", "City"] ).count()
g2=g1.add_suffix('_Count').reset_index()


df2=pd.DataFrame({'count' : df1.groupby( [ "Name", "City"] ).size()}).reset_index()

df3=pd.DataFrame({'count' : df1.groupby( [ "Name", "City"] ).size()})

#%% OLS Regression with groupby
M=500
df=pd.DataFrame({'FID': np.random.randint(low=1, high=3, size=500),
                 'MEAN':np.random.randn(M),
                 'Accum_Prcp':np.random.randn(M)*300,
                 'Accum_HDD':np.random.randn(M)*1000})

def group_ols(grps):
    results=[]
    for fid, grp in grps: # fid此处必须要有，是指index列。
        y=grp.loc[:,'MEAN']
        x=grp.loc[:,['Accum_Prcp', 'Accum_HDD']]
        result=sm.OLS(y,x ).fit()
        results.append((fid, result.summary()))
    return results
r_ols=group_ols(df.groupby(['FID']))
r_ols

#%% group predictions

# predict the value for each group for 01-10-2016.
df4=pd.DataFrame({'group':['A','A','A','A','A','A','B','B','B','C','C'],
                'date':['2016-1-2','2016-1-3','2016-1-4','2016-1-5','2016-1-6',
                        '2016-1-7','2016-1-2','2016-1-3','2016-1-4','2016-1-2','2016-1-3'],
                'value':[16,15,14,17,19,20,16,13,13,16,16]})

def model(df4, delta):
    y = df4[['value']].values
    X = df4[['date_delta']].values
    return np.squeeze(LinearRegression().fit(X, y).predict(delta))

def group_predictions(df, date):
    date = pd.to_datetime(date)
    df4.date = pd.to_datetime(df4.date)
    day = np.timedelta64(1, 'D')
    mn = df4.date.min()
    df4['date_delta'] = df4.date.sub(mn).div(day)
    dd = (date - mn) / day
    return df.groupby('group').apply(model, delta=dd)

group_predictions(df4, '2016-1-10')