Scribd
Upload
219 views17 pages

Hospital Patient Footfall Prediction

This document outlines a project to analyze and predict patient footfall at hospitals using time series forecasting models. The objectives are to apply forecasting models to predict future patient volumes in order to help with resource planning. The project uses a CRISP-DM methodology and analyzes over 1 million patient records to build Holt-Winters and Auto-ARIMA models. The models are deployed through a web-based interface that allows users to select patient types and view forecasts and error metrics.

Uploaded by

Mazhar Mahadzir
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
219 views17 pages

Hospital Patient Footfall Prediction

This document outlines a project to analyze and predict patient footfall at hospitals using time series forecasting models. The objectives are to apply forecasting models to predict future patient volumes in order to help with resource planning. The project uses a CRISP-DM methodology and analyzes over 1 million patient records to build Holt-Winters and Auto-ARIMA models. The models are deployed through a web-based interface that allows users to select patient types and view forecasts and error metrics.

Uploaded by

Mazhar Mahadzir
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

Hospital Patients footfall

Prediction
Project Goals

Objectives Constraints
• Analyse hospital patients footfall • Minimum number of datasets
• Applying forecasting model to predict patients • inconsistency of data
• missing data
footfalls in the future for planning and
maximizing allocation of healthcare resource to
meet the medical demands.

2
Business pipeline:

CRISP-DM Methodology

3
Technical Stacks

Application and Software Packages Used for This Project.

4
Project Architecture / Data Pipeline

5
Data Preparation
• Data used in this project was obtained from each respective
hospitals patient registration which has 16 vars (ID, Date-reg, Doctor
name, Old/New patient, Age/Age group, Gender, Hospital city and
location, Visit date, Cost, Service Items, Service Dept/Category,
Departments), and size of 1,032,509 observations of 3 months data
(Nov 2020- Dec 2021)

• Data cleansing
• Standardize data writing, spelling, data types, structure
• Data duplication removal
• Time series analysis – input/edit missing data due to holiday and
possible mis-entries by averaging same day data
• Training-Test Data Split by applying 80-20 (80% head as Training, and
20% tail as Test)

6
Model Building

7
Model Description

Algorithms used:
Time Series Prediction (Holt-Winters and Auto-Arima)
Prediction using : Auto-ARIMA

Descriptions:
Split the data into training (80%) & testing (20%)
Forecast whole set of time series data using Model and Comparing each accuracy

8
Outputs of the Model in Codes

screenshot
of model
codes

9
Training and testing data

original data

test data and


forecasting on
test data

compare

forecasting on
training data
original data
Forecasting 14 days footfall prediction
Comparison of Model Performance

12
Deployment Strategy

• Web Based

• Enterprise Based

13
Outputs of the Deployed Model

Outputs of
Universal Sentence
Encoder:

14
Output of the deployment

1. Side panel
2. Select patient type

3. Select number of
days you want to predict
Output of the model deployment
4. browse the csv file
6. Forecasting footfalls for the no of days
prediction selected

7. forecast plot

5. view the first 5 rows of data


8. view Mape and RMSE
value

You might also like