To deal with the problem of detecting the concentration of Fe2+ and Fe3+ in the goethite method to remove
iron process of a smelting production enterprises, on the basis of Outokumpu production equipment and technology, by
using improved least squares support vector machine to realize the nonlinear prediction for small sample and process neural
networks to realize the time cumulative effect of historical data series, an integrated prediction model is proposed based on
information entropy method. Simulation results show that, the integrated prediction model has good prediction performance
which satisfies the error request of the iron concentration in goethite method to remove iron process.