model training
Model training is a big part of Machine learning. it is important to ensure a proper division between training and evaluation efforts.
It is important to evaluation the model to estimate quality of its predictions for the data that the model has not been trained on.
bUT as a starting point your cannot check the accuracy of predictions for future instances as its supervised learning. so you need to use some of the data that you already know the answer for as a proxy for future data.
Instead of using the same data that was used for training to evaluate. A common strategy is to split all available labeled data into training set, validation set and test set often in 80:10:10 ratio or 70:15:15
Model Evaluation
After the model has interacted with unseen test data, we can deploy the model to production and monitor its to ensure business problem was indeed being addressed.
Its ability to more accurately predict skils, would reduce number of transfers a customer experienced. Thus resulting to a better customer experience. Model evaluation is used to verify that the model is performing accurately.
MODEL TUNING AND FEATURE ENGINEERING
Once we have evaluated our model and began the process of iterative tweaks to the model and our data. We can adjust how fast or slow the model was learning or taking to reach an optimal value.
then we move to feature engineering,
Feature engineering trying to answer questions like what was the time of a customer most recent orders, what was a customer most recent order....we feed these features into the model training aligorithm, it can only learn from exactly what we show it.
MODEL DEPLOYMENT
deploying the model, to solve the business needs and meet the expectations suh as directing customer to the correct agent the first time. Imagine if a company has a endless types of products, customer can be sent to a generalizt or even a wrong specialist, who will then figure what customer needs before sending them to agent with right skills... for a company handling millions of customer calls, this is inneffiecient and costs money and time.
customer calls get connected to..wrong department, non-technical support..then correct agent...
Top comments (0)