Predicting Volume and Composition of Municipal Solid Waste Based on ANN and ANFIS Methods and Correlation-Regression Analysis
Fragment książki (Rozdział w monografii)
MNiSW
50
Poziom II
| Status: | |
| Autorzy: | Dudar Igor N. , Yavorovska Olha V. , Zlepko Sergii M., Vinnichuk Alla P., Kisała Piotr, Shortanbayeva Aigul, Borankulova Gauhar |
| Dyscypliny: | |
| Aby zobaczyć szczegóły należy się zalogować. | |
| Wersja dokumentu: | Drukowana | Elektroniczna |
| Język: | angielski |
| Strony: | 13 - 24 |
| Efekt badań statutowych | NIE |
| Materiał konferencyjny: | NIE |
| Publikacja OA: | NIE |
| Abstrakty: | angielski |
| This chapter deals with the methods of predicting the volume and composition of municipal solid waste (MSW). After reviewing the existing methods of predicting the volume of MSW production, the authors classified four groups based on time series methods, deterministic correlations methods, GIS cluster analysis, and statistical learning theory. The suggested classification takes into account the morphological composition of MSW and the on-going changes in household behavior. The indirect and direct impact factors were determined with the formal-logistic method. Three mathematical models were compared to find an effective prediction model, namely, Artificial Neural Networks (ANN), Adaptive Neuro-Fuzzy Inference Systems, and correlation-regression models. The models were compared in order to find the most accurate one and describe measurement errors – the determination coefficient. The ANN and ANFIS models are significantly more precise than traditional correlation-regression models. The advantage of the models is that they are not as sensitive to non-standard input conditions as the correlation-regression model. However, ANN and ANFIS are more labor-intensive, requiring complex development and interpretation. We found that the ANFIS model outperforms the other two models in predictive accuracy at the beginning; however, the testing showed that it has a higher error margin than the ANN model. In addition, the learning process takes a long time for the ANFIS model, making it inapplicable in cases where a quick response is required. Therefore, the high accuracy of MSW prediction characterizing ANN makes it a valuable tool in determining strategies for optimization of MSW processing for a better circular economy. |
