ICOM 223 Sujoy 1

75 %
25 %
Information about ICOM 223 Sujoy 1
Education

Published on March 10, 2014

Author: TIIKMConferences

Source: authorstream.com

PowerPoint Presentation: Hosted by: Conference partners: FORECASTING FOOD DEMAND FOR HOSPITALS AND REDUCING WASTE Sujoy Roychowdhury FORECASTING FOOD DEMAND FOR HOSPITALS AND REDUCING WASTE : Sujoy Roychowdhury Alumnus- IIM, Bangalore FORECASTING FOOD DEMAND FOR HOSPITALS AND REDUCING WASTE BACKGROUND: BACKGROUND Food preparation in hospitals is an important but auxiliary part of their service Contributes to 4.5% of overall costs of medium-sized hospitals ( ~500 in-patients) Critical to reduce wastage and save on costs Food wastage is as high 50% of total waste in hospitals Most hospitals use “intuitive” forecasts for next days’ food demand Core in providing customer satisfaction to patients and relatives from both health and mental viewpoint Challenges of food demand estimation: Challenges of food demand estimation Large number of food items in each meal based on dietary requirements and constraints Hospitals in large cosmopolitan cities need to variety of items to cater to food preferences of its patients Patients’ physical and mental health conditions are such that they dislike all types of food Generally outsourced to catering services organisations and often appropriate historical data is not available Lack of skilled resources in scientific forecasting methods in such foodservice organisations Literature Review .. (1/2): Literature Review .. (1/2) Author Article Source Relevant Obs. Harris & Adam Forecasting Patient Tray census for hospitals Health Services Research, Winter 1975 pp 384-393 Mix of intuitive and statistical methods with the former being more prevalent Mattaso & Schalch Hospital Waste Management in Brazil: A Case Study Waste Management & Research Dec. 2001 pp 567-572 Food waste is 50% of total waste in hospitals Patil & Shekdar Health care waste management in India Journal of Environmental Management Oct 2001 pp 211-220 Proportion of food waste ~ 40-50% Literature Review .. (2/2): Literature Review .. (2/2) Author Article Source Relevant Obs. Engstrom & Carlsson-Kanyama Food losses in food service institutions: Examples from Sweden Food Policy June 2004 pp. 203-213 Classification of food waste Alam , Sujauddin , Iqbal, & Huda Healthcare waste characterization in Chittagong Medical Hospital, Bangladesh Waste Management & Research, Jun 2008 pp 291-296 Food wastage remains largest component of wastage in hospitals Goonan , Miranda, & Heather Getting a Taste for Food Waste … Journal of Academy of Nutrition & Dietics Jan 2014 pp. 63-71 Most hospitals do use electronic forecasting systems even now Miller, McCahon & Miller Foodservice forecasting using simple mathematical models Journal of Hospitality and Tourism Research, 1991 pp 43-58 Use of De-seasonalised exponential (DES) smoothing Classification of Food Waste: Classification of Food Waste Storage losses - which occur because of improper storage Preparation losses - which are mostly seeds, peel, etc. from fruits and vegetables Serving losses - which are what are left on serving dishes, and in canteens and bowls Leftovers - which are prepared food never served Plate Waste - which is what the diner leaves on the plate Focus of study on Serving losses and leftovers primarily Business Problem : Business Problem Study conducted for a hospital in South Bangalore Objective is to estimate breakfast food demand in the hospital - reduction of waste and ability to procure raw items in an appropriate basis Study consists of 28 different food items served at breakfast Quality experts and food service department has the opinion that it is dependent on occupancy (no. of hospitalized patients) during breakfast Likely to be dependent on quantity and food items of last few days since hospital occupancy is of 3.6 days Dataset: Dataset Breakfast occupancy and food demand for each of 28 items between 1 st October 2012 and 23 rd January, 2013 (115 days) This was daily manually collated data from menu cards given to patients 40 missing data points (1.2%) and 3 data points discarded for obviously wrong data Data corrected for these using multiple imputation method on SAS 29 th food item not considered as data was missing for whole of December Data from 1 st Oct 2012 to 31 st Dec 2012 used as estimation sample ( 92 days) and 1 st Jan 2013 to 23 rd January (23 days) as holdout sample Modelling done using SAS 9.2 TS Level 2M2 W32_VSPRO platform Food Items: Food Items Food items served in breakfast Boiled Egg Chutney Coffee Continental Breakfast Curd Custard Dosa Egg white omelette Fruits Dish Health drinks Idly Juice Milk North Indian Breakfast Oats Omelette Papaya Pongal Rava Rice Porridge Salad Sambar Soup Non-veg Tea Tomato soup Upma Veg Sandwich Veg soup Chicken Sandwich Sample Time Series Plot: Sample Time Series Plot Models of estimation chosen: Models of estimation chosen Model I - Regression with ARIMA Errors ………( 1a) ………( 1b ) Model II – Dynamic Panel Regression ….. (2) Model III - Transfer Function Model ….. (3)   Pros and Cons of Models: Pros and Cons of Models Pros Cons Model I Intuitive Interpretation of coefficients Somewhat difficult to estimate Does not consider lagged quantities , expressions Model II Estimation can be done in one step without detailed understanding of correlation plots and time series estimates Considers lagged values of dependent variables i.e. food item quantity Underlying theory of parameter and specification testing is very complex ( Arellano-Bond estimator) Model III Considers lagged values of independent as well as dependent variable Closest mapping to business description of problem Difficult to estimate Requirements for model acceptance: Requirements for model acceptance Each Food item will need to be modelled so that The MAPE <= 10% If MAPE >= 10% (because quantity is small) MAD <=3 Holdout Sample Metrics Summarized: Holdout Sample Metrics Summarized     Model I Model II Model III MAPE MAD MAPE MAD MAPE MAD Pass 18 3 13 3 20 5 Fail 10 7 15 12 8 3 Summary: Summary Model III has clearly better performance Model II has significantly poor performance Model I has medium performance Model III is recommended model because of Better performance on holdout sample Closest to business problem However Model III is difficult to maintain if re-estimation of parameters is required at future date Excel based estimation tool built for estimation of food demand with same parameters Shortcomings and future scope: Shortcomings and future scope Will need re-estimation of parameters for different hospital at different location Susceptible to changes in food menu – may affect other food items estimation if new items are introduced Future scope of work Extension to other businesses like foodservice, tourism etc. where there is a menu/catalogue driven ordering Handling of significant correlations between food items in model Q&A: Q&A

Add a comment

Related presentations

Related pages

ICOM 223 Sujoy 1 - slidesearch.net

ICOM 223 Sujoy 1 - slidesearch.net
Read more

Icom FL-223 SSB Filter | Icom Amateur Radio

Icom FL-223 9MHz 1.9kHz SSB Filter. For Icom amateur radios. In Stock and On Sale. Learn more... Need Help? Got Questions? CALL NOW! 1.855.289.2929.
Read more

ICOM R 224 Shambavi - slidesearch.net

ICOM R 224 Shambavi - slidesearch.net
Read more

ICOM FL 223 1 9KHZ SSB Narrow Filter 9MHZ IF 706 775 | eBay

Find best value and selection for your ICOM FL 223 1 9KHZ SSB Narrow Filter 9MHZ IF 706 775 search on eBay. World's leading marketplace.
Read more

icom fl-223 | eBay

Find great deals on eBay for icom fl-223 icom ut-106. Shop with confidence.
Read more

Sujoy Ghosh | LinkedIn

View Sujoy Ghosh’s professional profile on LinkedIn. LinkedIn is the world's largest business network, helping professionals like Sujoy Ghosh discover ...
Read more