for Business Intelligence-I need you write Part b Task B (section a-c)

Table of Contents 1

Executive Summary 2

Introduction 2

Justification of the choice of the series 2

Comments on the nature of the series 3

Factors likely to affect the nature of the time series 5

1. Political stability of Australia 5

2. Tourists sites availability 5

3. Visa Accessibility 5

4. Immigration policy of current government 5

5. Global Economy 5

Smoothen the series by seasonally adjusted data 6

Models provided to forecast the data 6

Holt-Winters Model 6

Exponential Smoothing Model 7

Choice of model 7

References 9

Appendix 9

Business forecasting techniques are used in this report to predict the number of tourists in Australia in the next 12 months. Public data on number of tourists is not directly available and short-term arrivals data is used. The time series showed both seasonal and trend components. The smoothing method used in this report include exponential smoothing method and Winters’ method. Winters’ method is chosen to predict the short-term arrivals as it shows better accuracy measures.

This report attempts to identify a time series that can be used in predicting the number of tourists in Australia from July-2017 to June-2018. The report applies times series smoothing techniques and also explains the rationale of using short-term arrivals to proxy the number of tourists in the future.

The report is broken down into several parts. The first part is to explain the reason to choose short-term visitor arrival. The second part is to investigate characteristics of the the times series. Time and seasonal factors are found. The third section of this report will try to smoothen the series and the second last section of the report will choose the appropriate model to make forecasts.

Time series refers to measurements collected at a successive equal interval over certain duration of time (measured in years) (Konar & Bhattacharya, 2017). Time series analysis can be applied in sales forecasting, stock market analysis, census analysis, yields projection or even economic forecasting. In this analysis, we are applying time series to study tourism growth rate in Australia, and possibly predict future forecast.

Time series data are usually used in forecasting future trends of events. The main components of time series are seasonal, cyclic, trend or even irregular time series data. In order to project trends of tourist arrivals in Australia, past records of arriving visitors are the most appropriate data that can be used to predict future expectation outcomes in the tourist industry (Box et al., 2016).

In this particular analysis, new arrivals at Australia are mostly people on vocational holidays, business trips or even honeymoon. These visits can be classified as seasonal since the measurement of the population variables are not constant and fluctuate with different interests of different individuals or groups. In order to predict future trends of visitors in Australia, historical figures are used to give a preliminary expected rate of tourist expectations in Australia in different months. It should however be noted that long-term visitors data is not suitable for forecast since persons on long-term visits may not necessarily be tourists in the Australian land (Leask, A 2016).

The time series plot of short-term visitor arrivals are used to predict the number of tourists in Australia in the coming year. The time series plot of the figure in the past 10 years are displayed as below. The source of the series is from Australian Bureau of Statistics (abs.gov.au)

Figure 1 Time series plot of Short-term visitor arrivals

The time series graph depicts a seasonal tend, with minimal cyclic valuation movements (Miah et al., 2016). It can be observed that a population of approximately 450000 visits in Australia between the months of January-February, 2006. At the end of the seasonal trend with cyclic valuation movements, the graph simply reflects a figure of approximately 720000 populations of new short-term visits (tourists) in Australia in the period of January, 2017.

Cyclic time series feature can also be found in the series. The directly proportional seasonal time series curve has oscillations vibrating up and down throughout the entire period between January-February, 2006 and early in the year 2017.

The graph is observed to have a slow or steady growth between the period January-February, 2006 and April, 2012 after which a rapid increase is observed between April, 2012 and early the year 2017.

Politics is a major factor that can influence tourist attraction in a given country. The nature of political aspects of a country can influence tourists’ visits either positively or negatively (Saarinen & Nepal, 2016). Stable political nature of Australia politics has driven to the increased visits of foreigners to Australia. This has in return produced very favorable tourists’ attraction sites in different regions of Australia.

Availability of physical features, attractive and beautiful features historical aspects of a country and many more spots are likely the main component that attracts visits to a certain country. Australia is one of the most beautiful countries and is surrounded by numerous physical features which attract visitors from all corners of the world. This has boosted the country tourism by double digits in the last two years.

Quick and easily accessibility of visa is another favorable aspect of movement that boosts travelling. Australia has accessible visa from the different countries, this has enhanced new visitors in the country from different countries. Visa accessibility and approval policy for this country has in return boosted the rate of new visits in Australia.

Immigration refers to the movement from one country to another. Favorable immigration policies in Australia has enhanced movement to-and-from the country. This has resulted to a growth in the tourism department of the Australians.

Australia has a very developed economy. Businessmen and other delegates’ from different country visit the land to hold business meetings and also learn more about the country (Saarinen & Nepal, 2016). The arrival of visitors is a boost to the country economy and also its tourism aspect.

The time series is smoothened by the method of Holt-Winters seasonal multiplicative model. The method consists of three smoothing equations, one for level, one for trend and one for seasonal component. We also need three smoothing parameters. In this case, we are looking for a forecast a monthly data, therefore, the period of seasonality used in this smoothing method is then 12. The three smoothing parameters are optimized to get the following set of parameters

The adjusted seasonal data using Holt-Winters Smoothing is shown as below

The forecasts are made based on exponential smoothing and Holt-Winters seasonal multiplicative model.

The prediction made by Holt-Winters seasonal multiplicative model with systematic parameters used in the previous smoothing part is summarized and tabulated as below

Table 1 Forecasts made by Holt-Winters Smoothing model

Another model that is used in this analysis is exponential smoothing model. The smoothing parameter used in the study is 0.2, which has been optimized by the Solver in Excel, using Root Mean Square Error (RMSE) of in-sample forecasts as selection criterion.

Table 2 Forecasts made by Exponential Smoothing model

To evaluate the performance of the model , we can examine the in-sample forecasting performance between two models. Two information metric will be used in the evaluation in this report, namely Mean Absolute Deviation (MAD) and Mean Absolute Percentage Error (MAPE). Both parameters are used to compare the accuracy of in-sample forecasts by these two models in order to make the choice.

The corresponding figures of both models are tabulated as below

Table 3 Measures of forecasting errors

From the table above, it can be found that Holt-Winters method performed better in in-sample forecasts in term of MAD. Therefore, this report will apply Holt-Winter multiplicative model as the forecasting model used to predict the number of tourists. The forecast can be plotted as below.

Box, G, Ljung, G, Reinsel, G, & Jenkins, G 2016, Time Series Analysis : Forecasting And Control, Hoboken, New Jersey: Wiley, Discovery eBooks, EBSCOhost, viewed 11 April 2017.

Konar, A, & Bhattacharya, D 2017, Time-Series Prediction And Applications : A Machine Intelligence Approach, Cham, Switzerland: Springer, Discovery eBooks, EBSCOhost, viewed 11 April 2017.

Leask, A 2016, 'Progress in tourism management: Visitor attraction management: A critical review of research 2009–2014', Tourism Management, 57, pp. 334-361, ScienceDirect, EBSCOhost, viewed 11 April 2017

Miah, S, Vu, H, Gammack, J, & McGrath, M 2016, 'A Big Data Analytics Method for Tourist Behaviour Analysis', Information & Management, ScienceDirect, EBSCOhost, viewed 11 April 2017.

Saarinen, J, & Nepal, S 2016, Political Ecology And Tourism, Basingstoke: Routledge, Discovery eBooks, EBSCOhost, viewed 11 April 2017.

Forecasting chart

10