Title: Improved PCA-based selective features and ICNN-bi-LSTM architecture for bio and non-biodegradable classification

Authors: S. Sasirekha; P. Kavitha

Addresses: Department of Computer Science, Vels Institute of Science, Technology and Advanced Studies, Chennai, Tamil Nadu, India ' Department of Computer Application-PG, Vels Institute of Science, Technology and Advanced Studies, Chennai, Tamil Nadu, India

Abstract: Biodegradable substances can naturally decay by breaking down the composite organic material into simpler components and can feed the soil with a variety of nutrients. These substances are less dangerous to the ecosystem since they are often harmless and do not negatively impact the environment over time. A novel method was developed for bio and non-biodegradable (B&NBD) waste classification, where, data augmentation was performed using shifting, rotating and shearing operations. Then, modified wiener filtering (MWF) was deployed during pre-processing. Further, grey level co-occurrence matrix (GLCM), Holo entropy, VGG16 and statistical features were extracted. To select the most informative features, improved principal component analysis (PCA) was exploited. The classification of B&NBD wastes was done using improved convolutional neural networks (ICNN) and bidirectional long-short-term memory (bi-LSTM). Further, an improved score-level fusion (ISLF) process was employed to fuse the outputs of ICNN and bi-LSTM.

Keywords: waste management; modified wiener filtering; improved PCA; improved CNN model; improved score-level fusion; ISLF.

DOI: 10.1504/IJEWM.2026.155920

International Journal of Environment and Waste Management, 2026 Vol.40 No.4, pp.384 - 408

Received: 19 Apr 2025
Accepted: 20 Jun 2025
Published online: 26 Aug 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article