Whenever the production process increases slowly the usual production function i.e., having multiplicative error term do not behave well for forecasting purpose. In this situation, Cobb-Douglas type production function with additive errors may be applicable. In this regard, the estimation procedure is to be complicated. Hence, non-linear estimation procedures are applicable. However, both Cobb-Douglas type function with multiplicative and additive errors are to be estimated here. Different model selection criteria show that Cobb-Douglas type production function with additive errors is more suitable for forecasting the industrial production in Bangladesh.