top of page
LAFEJ Softwares
LAFEJ has a team of researchers and students who develop software for various purposes. Depending on the application, this software is made available for use by the scientific community, both nationally and internationally.
Machine Learning Recovery Data (MALERD)
Machine learning for time series correction
Time series are essential for understanding physical phenomena, enabling mathematical modeling for future predictions of natural variables. However, obtaining them can be challenging due to equipment failures and maintenance costs, leading to data loss that hinders analysis. Researchers attempt to fill these gaps with plausible values, but inadequate methods can introduce bias. Techniques such as neighboring point averaging, maximum likelihood, and machine learning algorithms have been used to recover data, each with advantages and disadvantages.
This page will showcase the MALERD software developed specifically for recovering lost data. Machine Learning packages were used to recover geophysical time series data.
The solution
Several issues can lead to data loss, such as power surges, outages, equipment failures, and more. The solution was to create a machine learning-based model to solve this problem, as the model learns from past patterns and can predict future or missing data with minimal error. Figure 1 shows a graph containing data acquisition errors.
Figure 1: Example of a spurious phenomenon in the time series of the virtual height of layer F (h'F).
Regression model
A regression model is a statistical technique used to understand and predict the relationship between a dependent variable (or response) and one or more independent variables (or predictors). The goal is to create a mathematical function that best represents this relationship, allowing us to predict the values of the dependent variable based on the independent variables.
Figure 2 below shows the final result for the same Figure 1.
Figura 2: A mesma série temporal mostrada na Figura 1, mas já corrigida usando machine learning.
Application interface
Interface
Explanation for using the app:
Browse File: Browse for the file in the window that opens
Load Data: when clicking on the load data button, your file must already be selected. This is when the model will learn from the data and deliver the correction result.
The rest of the buttons are only for plotting the original graph (with errors) and the graph that has already been corrected. There is also the option to save data and the graph, and finally the development information and some more information for use.
Use the app
Download instructions
To use our application, follow a few steps:
1 - fill in the form fields
- why do we want this information: we want to have access to know who is using our application and where.
2- After the form is filled out you will be redirected to a media fire page, download it there.
3-the computer's operating system constantly detects the application as malicious. To reverse this situation, we are already working to obtain the digital certificate, but at the moment we do not have it yet. So when you click on download, a warning field will appear in the download, click on "keep downloading".
bottom of page