Build a algorithm which would allow to predict the correct postage charges.

Hi everyone, I hope you are well. We run an ecommerce business where we ship our goods around the country, including several regional and metro places. We have about 5000 products', where each product costs us differently to send to different customers around the country. So e.g product A would cost $5/kg to ship to Zone A (There are 16 total zones), while it would cost us $2/kg for product B. However, product B would cost $2/kg to ship to Zone A ,while it would cost us $5/kg for product B. Now, ideally we would want to charge the customer accurate shipping, however, shipping is not created on a product level. Shipping for the 16 zones is charged in something called a shipping template. A shipping template is where you define the cost of shipping for each zone. E.g: You can make a template where Zone A would cost $3/kg, Zone B would cost $4/kg ... Zone P would cost $2.5/kg. Each product is assigned a specific template (A product can only have one template). However, multiple products can be assigned to one template. Note: All products assigned to one template will charge the same shipping as defined in the template. Now, the main issue is we can not make 5000 templates, the max we can make is 10-15 templates. So, the 5000 products with their variable costs need to be divided into these 10-15 templates. So, predicting what cost to put in each template that best fits the products that are being assigned to it is the challenge. As these costs change frequently, an algorithm would be appreciated that we can use on the ever-changing data. Please, I hope this clarifies everything. Appreciate responses, or message me if you need further clarifications. Note: Please, only reply if you feel like you are able to attempt this. Thank you.