How does 90% average accuracy compare?
Model accuracy ranges between 0 and 100%. The closer to 100%, the more accurate the model and its predictions.
What does an average error of 10% mean?
Our measures are on a 100 point scale. So on average, your scores will be +/- 5 points from your “true” score. While there is some error in our estimations (as there are in all models), it’s a very small amount of error.
Examples Of How We Check Our Models
Identify the best type of model to run
There’s a number of different types of models we can run—ridge regression, lasso regression, elastic net, neural net, the list goes on. We build processes into our performance evaluation to make sure we’re using the best type of model.
Validate the model to make sure we’re not overfitting or underfitting the model.
We take a look at the model performance of various models to identify the model with both the lowest training error and lowest test error.
Calculate how the error in our model affects your scores
By looking at the average error, we know just how different your results could be if the model predicts inaccurately. A small range of values means our model is predicting very close to perfect.