Harnessing Artificial Intelligence for Efficient and Sustainable Garments: A Predictive Analysis of Deep Learning and Machine Learning from Selected Areas around Dhaka City

Abstract Garment industries produce a significant amount of materials and process waste, which has become an environmental and economical concern. Detecting or classifying those wastes at an initial period can be challenging Implementation of Artificial intelligence (AI) constitutes a crucial element in the prediction and classification of waste materials. Machine learning (ML) and deep learning (DL) can forecast waste quantities and differentiate various waste types based on multiple datasets. To forecast this problem, specific AI models have been applied. Artificial neural network (ANN), convoluted neural network (CNN) models such as, mobileNetV2, denseNet121, efficientNetB0, random forest (RF), gradient boosting (GB) were used as AI models The dataset underwent several preprocessing procedures. To evaluate the effectiveness of the AI model implementation, metrics including accuracy, F1-score, recall, precision, training and validation accuracy, training and validation loss, and MAE were employed. The result shows validation accuracies of MobileNetV2, improved MobileNetV2, DenseNet121, EfficientNetB0, improved EfficientNetB0 are 85.12%,81.83%,83.86%, 93.89%,94.67% respectively by utilizing data augmentation, hyperparameter tuning, adam optimization, softmax, ReLU activation function, transfer learning etc. Random forest gave accuracy = 81.72%, precision = 82.10% F1 score = 81.18% and gradient boosting accuracy = 81.19%, precision = 82.16% F1 score = 80.86% and R² of 99.78% for small dataset and 90.31% for big dataset had been shown by ANN and with improvement 95.17% has been showed by utilizing hyperparameter tuning, data scaling, adding more hidden layers etc. These suggested multi-AI models can contribute to predicting and detecting waste and provide valuable insights for the garments sector to a more sustainable and environmentally responsible industry.