LSTM Failure Forecasting from Historical Data
Budget: $10 – $30 USD
We have a dataset with information on monthly failures in the car department. You must predict monthly failures for X months in the future, where X will be a value to be entered in the code: 1 or 2 or 3...
This will be prediction 1 - prediction x months ahead of the car department's failures.
With the same input dataset, we will include a new attribute, the total number of failures for the entire company (all departments) and not just the failures of the car department.
The output will be the same as prediction 1, the difference is that now we will work with input 2 attributes (monthly failures of the car department + monthly failures of the company that encompasses all departments).
This will be prediction 2
The last prediction will be an evolution of prediction 2, where we will include one more attribute: the number of interventions that occurred in the car department. So we will have three input attributes (monthly failures in the car department + monthly failures in the company that encompasses all departments + number of interventions in the car department).
The 3rd prediction output will be the same as the 2nd and 1st predictions: it will always be just the projection of monthly failures for the next X months ahead.
Output always we need the yhat value (forecast), the lower and upper bounds. Also a graphic with theses values
Delivery: Jupyiter Notebook.
Input dataset will be provided.
It is important to have graphs that show the evolution of training and testing, proving that they are performing well.
Also inform validations with performance metrics in the code.
This will be prediction 1 - prediction x months ahead of the car department's failures.
With the same input dataset, we will include a new attribute, the total number of failures for the entire company (all departments) and not just the failures of the car department.
The output will be the same as prediction 1, the difference is that now we will work with input 2 attributes (monthly failures of the car department + monthly failures of the company that encompasses all departments).
This will be prediction 2
The last prediction will be an evolution of prediction 2, where we will include one more attribute: the number of interventions that occurred in the car department. So we will have three input attributes (monthly failures in the car department + monthly failures in the company that encompasses all departments + number of interventions in the car department).
The 3rd prediction output will be the same as the 2nd and 1st predictions: it will always be just the projection of monthly failures for the next X months ahead.
Output always we need the yhat value (forecast), the lower and upper bounds. Also a graphic with theses values
Delivery: Jupyiter Notebook.
Input dataset will be provided.
It is important to have graphs that show the evolution of training and testing, proving that they are performing well.
Also inform validations with performance metrics in the code.