This page is trying to help to develop a tool to estimate the cost of the plastics as a regression problem. The estimation of the part’s cost is currently done by the industrialization experts. With a detailed breakdown, in order to support the design to cost method with this project we want to use ML-oriented methods to estimate the cost of the parts based on the data records in hands to allow the non expert like a Designer, Cost Officer to estimate the cost during earlier phase of the project, using just visible features.

I also developed a platform on which the private usage of the model can be accessable.

The final cost includes these hidden costs:

- The Cost of the mold affected by depreciation: It would be calculated by the cost of the injection machine and the tool life-span.

- section Material cost

- Added value: the cost of labor and machinery during part forming, which depends on the cycle time, the type of press, the hourly rate and the location of production.

- Transport and logistics

bird overview of pre-processing step

In this project I was faicng an imbalanced regression problem and the lack of data, so some papars had been read to tackle this problem scientifically. you can see the papers below:

there papers lead me to the following ideas:

this is full result of (traditional and custom) methods reported by mean absolute error percentage MAPE criteria, including imported methods from papers and neural network:

LR = linear regression, XGB=XGBoost model, XGB-smote=XGB+SMOGN

errors below is in MAPE error system, not MSE!

As 25% in MAPE system is equivalent with 5% in mean square error

MSE

(in my case), the results can be regarded as "acceptable" according to their number of records.