optimal blending of 3 models

#1
I have 3 models that have produced losses based on historical industry-wide mix of business (property) and particular clients mix of business (property). I have actual losses for industry and actual losses for a particular client. The goal is to blend the 3 models for estimating any future losses.

Here is an example of the data using industry losses:
Event # Losses in Millions
Model-1 Model-2 Model-3 Actual
Event 1 5,897 3,960 4,725 5,088
Event 2 2,537 1,764 696 2,636
Event 3 957 1,057 1,326 1,785
Event 4 932 599 622 630


My first inclination is to create a blend (weighting) that minimizes the squared errors for industry & client.

thoughts?