| Abstract (Eng): | Time series classification consists of using machine learning to learn from a known group to classify a new dataset using time series data.
Time series can be used in many domains and recognisable by being a series of data points ordered by time.
The time series used in this paper are records of the speed to finish assembling tasks.
Those recordings are then used to estimate the operator's experience levels.
The goal of this paper is to show if it is not possible to perform this job better using machine learning models.
First a preprocessing step is performed to filter the dataset. The classification first happens by using simple models and is later extended to neural networks.
An analysis of the filtered dataset provides more insight in how the data is grouped.
The final results are inconclusive but it does show that the models are able to do classification to an extent.
Both approaches reach 60% accuracy on average while the best cases for neural networks can even reach almost 70%.
|
|---|