Modelo basado en machine learning para la predicción de variables empleadas en el secado de semilla en horno microondas
Seed drying involves reducing the moisture content at recommended levels, using techniques that do not deteriorate their viability, avoid deterioration, heating and infestation during storage. The present project is based on developing a model based on Machine Learning for predicting the variables u...
Shranjeno v:
| Glavni avtor: | |
|---|---|
| Drugi avtorji: | |
| Format: | bachelorThesis |
| Jezik: | spa |
| Izdano: |
2019
|
| Teme: | |
| Online dostop: | http://repositorio.utc.edu.ec/handle/27000/5340 |
| Oznake: |
Označite
Brez oznak, prvi označite!
|
Označite | Izvleček: | Seed drying involves reducing the moisture content at recommended levels, using techniques that do not deteriorate their viability, avoid deterioration, heating and infestation during storage. The present project is based on developing a model based on Machine Learning for predicting the variables used in seed drying, as there is no any model related to this type of research, therefore it is not possible to obtain a good efficiency in drying, which it invests time, economic and material resources. For the prediction development process, predictive techniques were used, such as: Neural Networks and Decision Trees, in addition, a methodology involving the experimental and prototype method was used. Additionally, the Python programming language was used with the Skylearn tool, this last tool allowed the use of the MLPRegressor neural network, which was calibrated with the following main characteristic; 5 input layers and 5 output layers were used. Likewise for the prediction with decision trees RandomForestRegressor was used, the same one that was calibrated under the main characteristic that in this case was the use of 10 estimators and a random state of 42. MLPRegressor reached the following predictions, for the drying time : 99.18%, energy consumption: 98.17% and germination rate: 78.18%, while with RandomForestRegressor the following predictions were reached, for the drying time: 99.18%, energy consumption: 98.17% and germination rate: 88.72%, The models generated were evaluated by coefficient of variation achieving 95.47%, therefore, the use of these tools allowed ob have optimal results in the prediction of the variables of drying time, energy consumption and germination rate, which help reduce, save time and resources. |
|---|