Scribd
Upload
715 views20 pages

Forecasting Demand For Food at Apollo Hospital: Submitted by Group 2

The document discusses forecasting food demand at Apollo Hospital in Bangalore. Currently, the hospital struggles to track food consumption and wastes a lot of food. They are looking for a forecasting system to estimate required food quantities to reduce costs. The summary analyzes food consumption data and develops forecasting models for high demand items like dosa, idli, etc. using techniques like exponential smoothing and ARIMA. Correlation analysis shows occupancy has a weak link to demand. The recommendation is to use an appropriate forecasting model along with daily occupancy data captured digitally. Key concerns are maintaining and scaling the model across hospital branches with varying local food preferences.

Uploaded by

Nikhil Shah
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
715 views20 pages

Forecasting Demand For Food at Apollo Hospital: Submitted by Group 2

The document discusses forecasting food demand at Apollo Hospital in Bangalore. Currently, the hospital struggles to track food consumption and wastes a lot of food. They are looking for a forecasting system to estimate required food quantities to reduce costs. The summary analyzes food consumption data and develops forecasting models for high demand items like dosa, idli, etc. using techniques like exponential smoothing and ARIMA. Correlation analysis shows occupancy has a weak link to demand. The recommendation is to use an appropriate forecasting model along with daily occupancy data captured digitally. Key concerns are maintaining and scaling the model across hospital branches with varying local food preferences.

Uploaded by

Nikhil Shah
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 20

Forecasting

Demand for Food at


Apollo Hospital
Submitted By Group 2
ASWIN GEORGE JACOB B2020011
OM BOROLE B2020012
SAURABH CHAVAN B2020013
DHRUV GARG B2020014
GAURAV PANJWANI B2020016
SUJAY GOVEKAR B2020017
Scenario
Demand Constraints
Location: Apollo Hospital, Bangalore
No tracking of food
 The food distributed among
consumption patients is dependent on their
The Hospital Management is struggling to figure respective food diets and
out the amount of food that has been consumed restrictions from the dieticians.
by the in-patients.
 Patients would be likely to order
same food for their complete
hospital

Food Demand on: Need Demand Forecasting


 No. of inpatients currently being Looking for a system to estimate
treated in the hospital the required quantity of food and
 Hospital staff and caretakers beverages for now and has the
 Attendants of the patients Huge wastage of scope of scaling up in future.

food and beverages Major objective to reduce the


produced overall costs of hospital
Case Questions
&
Answers
1. Apollo serves 120 food items from its kitchen. Do you think that Apollo should develop
forecasting models for all 120 items?

It is Not Necessary
It is not necessary for Apollo to develop forecasting model for all the 120 food items. It
is enough if we develop forecasting models for groups of items as the forecasting
accuracy is better.

Apollo
Cafeteria
For High Value & High Consumption Items
The data for all 120 items must be recorded and forecasting must
be done for high value and high consumption item.

For Less Demanded Items


For some of the items, the historical demand registered is of very
less quantity, that shows no appreciable trend. In such cases, it is
better to take average of demand instead of spending time
developing a forecasting model
2. Explore the data provided in the supplement document. What inferences can be
obtained from the data?

Demand for Complementary Dishes


Comparing the mean, median and mode for all food items we find that chutney and
sambar have the same values for above parameter. This is because they are
complementary dishes.

Apollo
Maximum Wastage is in Dosa
If we compare the difference between maximum food prepared
Cafeteria
and minimum food that needs to be prepared, we can see Dosa
having 121%

Maximum wastage is in Dosa followed by Idli, North Indian B/f,


omelette, continental B/f and chutney and sambar

Consumption Vs Occupant
Also from the plot of consumption VS Occupant we see that till
190 units data is homogenous and after that data becomes
heteroscedastic
3. Develop an appropriate forecasting model for each of the chosen food items.

No Seasonality in Dataset Apollo


Because there is no seasonality present in the dataset, we have to Cafeteria
pick between Single Exponential Smoothing and Auto ARIMA.

The errors for all the models were calculated. MAPE and RMSE were
used to compare the models for each food item.
For Dosa
For Idly
For Chutney
For Sambar
For Continental B/F
For North Indian B/F
For Omelette
4. Discuss the results of the models developed in question 3 and its appropriateness for
use by Apollo Hospitals.

For Dosa

For Idly

For Chutney

For Sambar
4. Discuss the results of the models developed in question 3 and its appropriateness for
use by Apollo Hospitals.

For Continental B/F

For North Indian B/F

For Omelette
5. Dr. Rao has suggested that the food demand is related to the occupancy. Is there
statistical evidence for this belief?

Dr. Rao suggestion is Invalid

While calculating correlation between the individual food items demand and the occupancy, it is found that there is only weak
relationship between the two:

Correlation Idly Dosa Chutney Sambar Continental B/F North Indian B/F Omelette

BKFST_OCCUP 12% -6% -2% -1% -22% 11% 16%

The strongest relationship among the items is a negative relationship between occupancy and continental breakfast which
means that as occupancy increases, the number of people ordering continental breakfast shows a decline.

To check if there’s a causal relationship between these two variables, the R2 value has to be statistically significant. This is
not the case as below table of R2 values (quadratic regression model) for each of the food items with occupancy proves:

R2 Idly Dosa Chutney Sambar Continental B/F North Indian B/F Omelette

BKFST_OCCUP 1% 2% 0% 0% 6% 1% 4%
6. What will be your recommendation on the use of daily occupancy in the forecasting
model?

Recommendation

Daily Occupancy will aid in better forecasting actual


consumption based on actual hospital footfall.
The hindrance to this was readily availability of the data in
the desired format.
Most of the data were in paper format, in-order to
overcome this difficulties one proper format let say
updating the data digitally using tablets or mobile devices
would be helpful in the forecasting model.
Also in the model there was no seasonality observed, we
can go with the daily occupancy.
7. Suggest suitable action for Apollo hospitals going forward for a proper estimation of
food demand.

Action Plan
Using the appropriate forecasting models – Arima Forecasting Model
as verified above can be a major contributor to lessen the food and
beverage wastage in Apollo Hospital
8. Comment about Dr. Rao's concerns on the maintenance and scalability of the model

• Dr.Rao knew that developing the model is just


a head start and he knew that implementing
significant improvements in process was
more challenging than building forecasting
models.
• And as he was about to scale up the hospitals
as well as open few other branches, the main
concern would be, will the same model work
all the time for all the branches. Because in
India, stable food changes almost after every
next town, village or city. So, considering this
the existing model has to go through lot of
changes.
THANK YOU

You might also like