ignavinuales / predictive_maintenance

Predictive Maintenance (Binary and Multiclass Classification)

Brief Description

This dataset reflects real predictive maintenance encountered in the industry with measurements from real equipment. The features description is taken directly from the dataset source.

Objectives:

1. Finding insights by doing an Explanatory Data Analysis (EDA).
2. Developing two ML models.
   • One for binary classification to predict Failure or No Failure.
   • One for multiclass classification to predict Type of Failure.
3. Exploring some insights from the results of the models.

Challenges:

• Dealing with highly imbalanced data.

The six features are:

• Type: the quality of the product, consisting of a letter L, M, or H. Meaning low, medium, and high, respectively.
• Air temperature [K]: generated using a random walk process later normalized to a standard deviation of 2 K around 300 K.
• Process temperature [K]: generated using a random walk process normalized to a standard deviation of 1 K, added to the air temperature
  plus 10 K.
• Rotational speed [rpm]: calculated from power of 2860 W, overlaid with a normally distributed noise.
• Torque [Nm]: torque values are normally distributed around 40 Nm with an σ = 10 Nm and no negative values.
• Tool wear [min]: The quality variants H/M/L add 5/3/2 minutes of tool wear to the used tool in the process.

The targets are:

• Target: failure or no failure (to perform binary classification).
• Failure Type: type of failure (to perform multiclass classification).

It also includes the following information, which is not useful for building the models:

• UID: unique identifier ranging from 1 to 10000.
• ProductID: the id of the product.

Dataset source: https://www.kaggle.com/datasets/shivamb/machine-predictive-maintenance-classification

Confusion Matrices