Forecasting number of work-related injuries time series with Box-Jenkins Models for registered insured in SSIO between 2000 and 2010 in Iran

ABSTRACT

Controlling occurrence of accidents in work place has been an interesting subject in all countries worldwide. Financial consequences of these accidents and their economic losses imposed on the involved companies is only one of the insignificant aspects of such damages and when the non-economic but intangible losses to the society are taken into consideration ,these economic damages will be marginalized. Purpose of this research is fitting the box-Jenkins model to time series of total number of accidents in work place and estimation of series1 missing values during fitting of this model. This longitudinal [time series] study, intends to model, estimate and forecast time series of total number of work place accidents for the insured people between 2000 and 2010 in Iran. Spline non-parametric regression methods to find the best interpolation and estimation of the series' missing value as well as box-Jenkins prediction method to find the best prediction on series have been used. Results: Smoothing spline method using some adjusts for seasonal time series in order to estimate missing values, shows better performance on the series. Then, suitable box-Jenkins model, SARIMA [0,1,1] x [0,1,1][12], was fitted to the series. Goodness of fit criterion of model, AIC, for prediction of the time series was equal to 1050.833. Estimation of the missing data in time series with regard to characteristics such as long period of missing data and exposure of values in tandem without any reliable middle points, is an appropriate method in order to leave situation of such gaps in the data. Looking at the goodness of fit of the model can be said that the error estimation and prediction model is rather low and hence, the model is well fitted to the series, Therefore, box-Jenkins model will be reliable for fitting to similar data