Incineration of waste materials has become a common way to handle a worldwide increase of produced waste materials. The problem with waste as fuel is that the material is not homogeneous and include a mixture of fossils and renewables. The fossil part ends up in an emission of fossil carbon dioxide in a combustion process, which is included in the emissions trading system. However, since waste material varies a lot in composition depending on the time of the year, origins, etc. it is preferable to have a method for real-time measurement of the fossil share of combusted waste. No real-time measurement technologies are available today which is the reason to investigate if near-infrared (NIR) spectroscopy could be a potential solution. An artificial mixture of refused derived fuel has been used to investigate the possibilities of NIR for prediction of the fossil share in waste material. The fossil share is assumed to be equal to the content of plastic material with an origin of oil products. Mixtures with different plastic content are scanned by the NIR instrument to obtain individual absorption spectra. A Partial least square (PLS) regression model is created on measured spectra and known content of plastics. The best model for the prediction on new spectral data using one of four measured replicates is a PLS model preprocessed with Savitzky-Golay smoothing that gives an R-square value of 0,782. If the prediction is done, depending on a delimitated wavenumber interval and an average of all four replicates is the best model instead of a PLS model pre-processed with standard normal variate without seven outliers that have an R-square value of 0,81. R-square value is the coefficient of determination which has been used to figure out the best model. An R-square value above 0,65 are recommended for process modelling, where 1 is the highest possible value.