Literature Review

This section provides a state-of-the-art ARIMA models as used in the literature for predictive analytics. Table 13.1 presents the comparative analysis of various works.

Findings

From this study, it is clear that efficient and effective dengue forecast tool with higher accuracy a pressing need in order to control and prevent such outbreaks all over the world. This study reveals that time series model using meteorological data has been successful aimed to forecast of specific transmissible diseases.

Methods and Materials

Study Area

In this research, two datasets are taken to predict the dengue incidence rate using meteorological data in Tamil Nadu. Dataset 1 contains weekly dengue cases data for the periods 1990-2009 of Tamil Nadu. Dataset 2 contains weekly minimum temperatures, maximum temperatures, average temperatures, rainfall, and precipitation for the period of 1990-2009 in Tamil Nadu District (Yong-Su Kwon 2015)

Snapshot for Dataset

Dataset 1

week start date

tota l_cases

4/30/1990

4

5/7/1990

5

5/14/1990

4

5/21/1990

3

5/28/1990

6

6/4/1990

2

6/11/1990

4

6/18/1990

5

6/25/1990

10

7/2/1990

6

7/9/1990

8

Comparative Analysis

Technique

Description

Dataset

Recommendation/Prediction

Exponential smoothing ARIMA

STL decomposition forecasts trend and seasonal patterns in a better way.

(Chiung Ching Ho 2015)

Malaysian Open Data Portal

To study the impact of the out-of-sample data on the models and integrate correlation studies with meteorological data.

Prediction: Dengue fever outbreaks in Malaysia

Trend of morbidity and mortality

Tamil Nadu has been witnessing a fall in the morbidity trend over the past five years, and dengue deaths since 2015. Nevertheless, cases of dengue morbidity are increasing and median mortality is rising from 2012 to 2016, nationally.

Morbidity and dengue-related mortality at Puducherry and Tamil Nadu, India (2012-2016).

(Sahanaa and Mishra 2018)

This work had assessed the Dengue in Puducherry and Tamil Nadu and proposed the model to reduce and eradicate the dengue outburst

Prediction: early detection of the dengue outbreak, in fact prediction of dengue (Sahanaa and Mishra 2018)

Statistical analysis and modeling: Time series analysis, Bivariate analysis, Multivariate analysis.

Dengue cases in Noumea were basically driven by climate during the last forty years. (Descloux 2012)

Epidemiological data:

January 1971- December 2010 Meteorological data:

January 1971 to December 2010 Entomological surveillance data: since 1997

(Earnest 2012) (Elodie Descloux 2012) Prediction: Climate-based Dengue Epidemic Models for Understanding and Forecasting

support vector machine (SVM), classification and regression tree (CART), and random forest (RF).

To examine the spatial and temporal vanations in the frequency of urban mosquitoes and the relationships with meteorological and habitat conditions such as type of land use.

Mosquito data collected from 2011 to 2012 at 12 locations, and environmental data (Yong-Su Kwon 2015).

Applied for the efficient control of urban mosquitoes.

Prediction: Mosquito Occurrence (Yong-Su Kwon 2015).

(Continued)

Comparative Analysis

Technique

Description

Dataset

Recommendation/Prediction

Auto Regressive Integrated Moving Average (ARIMA) Model.

The trend in forecast dengue cases for the years 2018 to 2025 shows that there is a stable growth of Dengue cases, which is of serious concern.

Dengue cases in Tamil Nadu 1997-2017.

(Karnaboopathy and Venkatesan 2018)

To avoid the disease from becoming endemic, new interventions with increased intensity of existing interventions and help from the international community together with the WHO are essential in order to stop the epidemics.

Prediction: Number of cases till December 2025

Seasonal Autoregressive Integrated Moving Average (SARIMA) models

To model the monthly number of dengue fever (DF) cases in Dhaka, Bangladesh, and forecast the dengue incidence using time series analysis (Zamil Choudhurya and Banu 2008).

Dengue fever (DF) cases in Dhaka Bangladesh monthly data January 2000 to October 2007 (M.A.H. Zamil Choudhurya and Banu, 2008).

Separate modelling approaches for DF, DHF and DSS would provide better information to policy-makers and planners Prediction: forecast for the period. November 2007 to December 2008.

Dataset 2

w e e k_sta rt_d at€

TMAX

TMIN

TAVG

RAINFALL

PRCP

4/30/1990

28.78571

21.74286

25.25714

7.042857

1.428571

5/7/1990

29.11429

21.58571

25.32857

7.528571

3.014286

5/14/1990

27.92857

21.65714

24.78571

6.271429

7.314286

5/21/1990

28.4

21.1

24.72857

7.3

3.985714

5/28/1990

28.32857

21.28571

24.82857

7.042857

2.242857

6/4/1990

28.17143

20.71429

24.45714

7.457143

4.285714

6/11/1990

28.41429

21.44286

24.95714

6.971429

2.3

6/18/1990

27.77143

21.12857

24.47143

6.642857

2.571429

6/25/1990

28.48571

21.6

25.04286

6.885714

0.657143

7/2/1990

27.85714

21.28571

24.57143

6.571429

4.9

7/9/1990

27.38571

21.65714

24.52857

5.728571

3.271429

Proposed Model

Time series analysis Brockwell (2013) may be classified as linear and nonlinear (Dhamodharavadhani Rathipriya2020b). Extremely specific techniques are used for the study of time series ARIMA, such as the Box-Jenkins multivariate and Holt winter exponential smoothing (single, double, and triple). ARIMA models are traditional forecasting models that require historical empirical data as evidence to make predictions (Dhamodharavadhani and Rathipriya 2019). This model is a simple statistical framework that can be used as the basis for mathematical models. These three order variables (a, b, c) describe the process of fitting the ARIMA model to the Box-Jenkins system (Sahanaa and Mishra 2018).

The Figure 13.2 shows the methodology of time series forecasting using ARIMA model. The first step of the ARIMA model is preprocessing the data. Time series data are plotted and its patterns and irregularities are examined. Next, the outliers and missing values are removed. The second step is to decompose the data and then to stationary the series and after that to calculate the autocorrelation by choosing the model order based on the process to fit the ARIMA model, then to evaluate and improve the model. The following Table 13.2 describes the steps to fit the ARIMA to a dengue forecasting model and its mathematical format.

Estimate and Develop the Model

The prediction is iterated in two ways to enhance the forecast. One is to add the seasonal variable to be removed earlier. Another solution would be to require (А, В, C) components to be included in the model; refitting the model on the same results, so that its silent ability is determined by a particular seasonal trend in the sequence, with the seasonal aspect defined in AR (1) (Dalinina 2017). The parameters (a, b, c) have also been modified to include the seasonal variable. Once the same process of analyzing the residual model has been done, the ACF/PACF plots change the structure as appropriate. For

FIGURE 13.2

Workflow of proposed model.

example, finding the same higher-order evidence is present in auto correlations with lag 7, which indicates a higher-order component may be required.

 
Source
< Prev   CONTENTS   Source   Next >