You work for a social media company and are tasked with detecting whether posted images contain cars. You have developed an object detection neural network and deployed a model version to AI Platform Prediction for evaluation. For evaluation purposes, you attached an evaluation job to the model version before deployment. Each training example is classified within exactly one class. During evaluation, you observe that the model's precision is below the threshold required by your business standards. Knowing that there is a trade-off between precision and recall, how should you adjust the model's final layer softmax threshold to improve the precision of your model?