It is necessary to create a classification machine learning algorithm to determine plateaus in time-temperature data.
The data contains three parameters - time, temperature, zones.
Zones - an indicator of the plateau (0 - plateau, 1 - heating, 2 - cooling).
[In some experiments, the "plateau" may be at a 45° angle, so machine learning should be trained on manually labeled data, not mathematics to determine a "flat" section]
See examples of graphs based on the data, with the plateau marked by an oval.
The target accuracy for predicting the zones parameter is 90-95%.
Bayesian optimization of hyperparameters should be used (this is the technical specification).
Attached are the time-temperature input data in Excel:
training data - 4 experiments on 4 sheets in one file
test data - another 1 experiment in a separate file
https://docs.google.com/spreadsheets/d/1v3Tj_lhWme4-qF7wkuSNWIP-NEOtvxN2/edit?usp=drive_link&ouid=117387964102202863857&rtpof=true&sd=true, https://docs.google.com/spreadsheets/d/1o6GJZOsriIM2gZiB-de2OQsGMPMUuWOr/edit?usp=drive_link&ouid=117387964102202863857&rtpof=true&sd=true
1) It is necessary to create a classification machine learning algorithm to determine plateaus in time-temperature data.
The data contains 4 parameters - time, temperature, zones, (sensor)type
Time – time
Temperature – temperature
Zones - indicator of the plateau (0 - plateau, 1 - heating, 2 - cooling).
(sensor)type – type of sensor (there can be about 10 types of sensors in total)
See examples of graphs based on the data, with plateaus marked by an oval.
2) The target accuracy for predicting the zones parameter is 90-95%.
3) Bayesian optimization of hyperparameters should be used (this is the technical specification).
4) A brief description of the methods used in the code should be prepared, explaining why these methods were chosen (for example, why RandomForestClassifier was selected). If alternative methods can be applied, they should also be mentioned).
[Approximately 5 experiments will be available for each type of sensor]
[In some types of sensors, the "plateau" may be at an angle >30°, so machine learning should be trained on manually labeled data, rather than using mathematics to determine a "flat" section]
I see 2 options for implementing machine learning:
- one model for all types of sensors, using the parameter (sensor)type to differentiate types (this is the preferred option if it does not compromise the accuracy of plateau prediction).
- a separate machine learning model for each type of sensor, with each model trained separately on its own type of sensor
The task is not to obtain a file with a trained machine learning model now, but to obtain a method for training the model/models. So that it can be stated that certain methods were used and a certain result was achieved. This can be considered a scientific work.
Input data time-temperature in excel:
1) Sensor type – zn
training data - 4 experiments on 4 sheets in one file:
test data - another experiment in a separate file:
2) Sensor type – sn
[another 4-5 experiments will be available soon, similar to zn]
3) Sensor type – in
[another 4-5 experiments will be available soon, similar to zn]
Output data:
- file with the trained model
- accuracy_score and other standard metrics in the console
- accuracy_score and other standard metrics in an excel file on the 1st sheet, on the 2nd sheet test data + predicted data,
Example:
https://docs.google.com/spreadsheets/d/1iG_uskzq7LVYArw0ObljklQXWZfuQYO7/edit?usp=sharing&ouid=117387964102202863857&rtpof=true&sd=true
Log in or register, to view the original
Budget: 10000 UAH Deadline: 5 days
Good afternoon. I am ready to complete the task.
I have extensive experience in developing AI projects with machine learning.
My portfolio: 
Freelancehunt
I will develop a machine learning algorithm to determine plateaus in time/temperature data. I will implement training on manually labeled data. I will use Bayesian optimization of hyperparameters.
Write to me, we will discuss the details and I will start working.
Budget: 5000 UAH Deadline: 2 days
Hello, I am interested in the project and can complete it according to the specifications.
Budget: 5800 UAH Deadline: 7 days
Good day,
We will do everything clearly and qualitatively,
Check the portfolio for a better understanding of my experience.
