Material and Cost estimation of a Customized Product based on the Customer's description

Many companies develop customized products for their customers. The customer, with the assistance of the salesperson, will typically provide a description of the product and the company must then use that description to determine the type and quantities of materials required to produce the product. This information is then used to derive the cost of production from which a price can be determined. We use Machine Learning techniques to automate this process. The product description is analyzed using Natural Language Processing to extract the relevant information. This information, along with other attributes, are then fed into a Deep Neural Network (DNN). The DNN has an output for each possible component of the product with the output value equal to the quantity of that component required for the product. We illustrate the approach with a dataset taken from a company that builds electrical distribution boards. Each distribution board must be customized for the customer and so the accurate determination of the components and their quantities is vital in determining an appropriate price. Note that, the component list (called the Bill of Materials or BOM) also helps determine the processing required and this too affects the production cost. We illustrate the effectiveness of our approach with the data obtained from this factory.