Future demand at CCG level#

#turn warnings off to keep notebook tidy
import warnings
warnings.filterwarnings('ignore')

Run forecasting notebook#

%run ./stacked_forecast.ipynb

0.4519048888936743
0.4227235191823048
0.7115831805529126
0.7596144008785489

(1618, 14)
_images/ccg_forecast_3_2.png _images/ccg_forecast_3_3.png 
Combined training score: 0.8795260326236356
_images/ccg_forecast_3_5.png 
results
ccg month 2019 2020 2021 2022 2023 2024 2025 2026 2027
0 00Q Jan 15570.0 15382.955000 15098.926198 15100.848326 15101.859973 15874.355786 15869.250671 15863.294703 15856.381527
1 00Q Feb 14493.0 15175.861940 14891.747074 14893.642828 14894.640593 15667.169722 15662.131237 15656.253004 15649.430055
2 00Q Mar 15748.0 15175.861940 14891.747074 14893.642828 14894.640593 15667.169722 15662.131237 15656.253004 15649.430055
3 00Q Apr 15929.0 15252.708754 14968.625824 14970.531365 14971.534281 15744.051047 15738.987837 15733.080759 15726.224329
4 00Q May 16180.0 15321.412964 15037.358585 15039.272876 15040.280398 15812.786111 15807.700797 15801.767930 15794.881566
... ... ... ... ... ... ... ... ... ... ... ...
907 X2C4Y Aug 17191.0 16211.368152 16259.624026 16306.118469 16349.457014 16389.639662 16427.547128 16464.463788 16499.839197
908 X2C4Y Sep 17115.0 15858.666878 15905.872877 15951.355767 15993.751421 16033.059838 16070.142574 16106.256060 16140.861827
909 X2C4Y Oct 17229.0 15567.474340 15613.813555 15658.461301 15700.078497 15738.665144 15775.066976 15810.517357 15844.487702
910 X2C4Y Nov 17153.0 15474.697068 15520.760116 15565.141775 15606.510946 15644.867628 15681.052517 15716.291624 15750.059516
911 X2C4Y Dec 17813.0 15830.887182 15878.010490 15923.413707 15965.735096 16004.974657 16041.992434 16078.042661 16112.587808

812 rows × 11 columns

Plot for CCG#

results.ccg.unique()

array(['00Q', '00R', '00T', '01E', '01G', '01J', '01K', '01V', '01W',
       '01Y', '02E', '02H', '02P', '02X', '03F', '03H', '03L', '03N',
       '03Q', '03R', '04C', '05W', '06H', '06K', '06N', '06Q', '06T',
       '07H', '07K', '09D', '10Q', '10R', '11M', '11N', '11X', '12F',
       '13T', '14L', '14Y', '15A', '15C', '15E', '15F', '15M', '15N',
       '16C', '18C', '26A', '27D', '36J', '36L', '42D', '52R', '70F',
       '71E', '72Q', '91Q', '92A', '92G', '93C', '97R', '99A', '99C',
       '99E', '99G', 'A3A8R', 'B2M3M', 'D2P2L', 'D4U1Y', 'D9Y0V', 'M1J4Y',
       'M2L0M', 'W2U3Z', 'X2C4Y'], dtype=object)

26A#

ccg = '26A'

fig,ax = plt.subplots(figsize=(8,5))

scenarios = ['Population growth','111 capacity increase',
            'Ambulance capacity increase','GP capacity increase', 'Health of 2019']

for i,results in enumerate(scenario_results):
    
    if i==0:
        
        continue
        
    results_ccg = results.loc[results.ccg==ccg]
    
    points = sum_by_month(results_ccg)
    
    points_series = pd.Series(points)
    
    plt.plot(np.arange(-12, 96),
             points_series.rolling(window=4).mean().to_list()[:], label = f'{scenarios[i]}')
    
    
points = sum_by_month(scenario_results[0].loc[scenario_results[0].ccg==ccg])

points_series = pd.Series(points)
    
plt.plot(np.arange(-12, 96),
         points_series.rolling(window=4).mean().to_list()[:], 'g--', label = f'{scenarios[0]}')

    
y = np.arange(23500,28000,1000)    
#plt.plot(12*np.ones(len(y)),y, 'k--')


plt.legend(loc = 'best', fontsize=12)
plt.ylabel('Monthly ED attendances ', fontsize=14)
plt.xlabel('Months since Baseline', fontsize=14)
plt.xlim(0,48)

start, end = ax.get_xlim()
ax.xaxis.set_ticks(np.arange(-12, 50, 6))

plt.tight_layout()

plt.show()
_images/ccg_forecast_9_0.png

00Q#

ccg = '00Q'

fig,ax = plt.subplots(figsize=(8,5))

scenarios = ['Population growth','111 capacity increase',
            'Ambulance capacity increase','GP capacity increase', 'Health of 2019']

for i,results in enumerate(scenario_results):
    
    if i==0:
        
        continue
        
    results_ccg = results.loc[results.ccg==ccg]
    
    points = sum_by_month(results_ccg)
    
    points_series = pd.Series(points)
    
    plt.plot(np.arange(-12, 96),
             points_series.rolling(window=4).mean().to_list()[:], label = f'{scenarios[i]}')
    
    
points = sum_by_month(scenario_results[0].loc[scenario_results[0].ccg==ccg])

points_series = pd.Series(points)
    
plt.plot(np.arange(-12, 96),
         points_series.rolling(window=4).mean().to_list()[:], 'g--', label = f'{scenarios[0]}')

    
y = np.arange(23500,28000,1000)    
#plt.plot(12*np.ones(len(y)),y, 'k--')


plt.legend(loc = 'best', fontsize=12)
plt.ylabel('Monthly ED attendances ', fontsize=14)
plt.xlabel('Months since Baseline', fontsize=14)
plt.xlim(0,48)

start, end = ax.get_xlim()
ax.xaxis.set_ticks(np.arange(-12, 50, 6))

plt.tight_layout()

plt.show()
_images/ccg_forecast_12_0.png

72Q#

ccg = '72Q'

fig,ax = plt.subplots(figsize=(8,5))

scenarios = ['Population growth','111 capacity increase',
            'Ambulance capacity increase','GP capacity increase', 'Health of 2019']

for i,results in enumerate(scenario_results):
    
    if i==0:
        
        continue
        
    results_ccg = results.loc[results.ccg==ccg]
    
    points = sum_by_month(results_ccg)
    
    points_series = pd.Series(points)
    
    plt.plot(np.arange(-12, 96),
             points_series.rolling(window=4).mean().to_list()[:], label = f'{scenarios[i]}')
    
    
points = sum_by_month(scenario_results[0].loc[scenario_results[0].ccg==ccg])

points_series = pd.Series(points)
    
plt.plot(np.arange(-12, 96),
         points_series.rolling(window=4).mean().to_list()[:], 'g--', label = f'{scenarios[0]}')

    
y = np.arange(23500,28000,1000)    
#plt.plot(12*np.ones(len(y)),y, 'k--')


plt.legend(loc = 'best', fontsize=12)
plt.ylabel('Monthly ED attendances ', fontsize=14)
plt.xlabel('Months since Baseline', fontsize=14)
plt.xlim(0,48)
plt.ylim(78000,86000)

start, end = ax.get_xlim()
ax.xaxis.set_ticks(np.arange(-12, 50, 6))

plt.tight_layout()

plt.show()
_images/ccg_forecast_15_0.png

Summary#

It is clear that, when looking at individual CCGs, the forecasts can vary significantly from the mean results presented on the previous page.

This demonstrates that for each CCG the model is learning different relationships between the different variables.

Forecasts at the CCG level would be more accurate if local data was fed into the model