EN
The paper analyses a time series of the daily number of patients visiting an orthopedics clinic. A learning set was chosen to ensure the homogeneity of series variance and three forecasting models were built. The regression model consists of the linear trend, dummy variables describing weekly and yearly seasonal components and the auto-regression of the residuals. ARIMA, with non-seasonal and seasonal differencing, contains only the components of the first order moving average. The exponential smoothing model covers the linear trend and weekly harmonic component. The residuals from all three models very closely follow normal distribution. Forecasts have been compared with actual data for a monthly test period and all models allow us to forecast the number of patients, with a mean square error of 9 people. Exponential smoothing appears to have the lowest MAPE for the test period.