Canara Engineering College

Chikmagalur district of Karnataka state, situated within the boundaries of the Western Ghats is highly susceptible to landslides, especially during the monsoon season. Despite the recurring nature of these slope failures, limited research has been conducted to assess and mitigate the risk of landslides in the region. Existing studies often lack a comprehensive analysis of the triggering elements and rely on basic machine learning (ML) techniques, even though there are several advanced techniques that are being adopted across the world. A comprehensive dataset of the study area was prepared by integrating twenty different Landslide Conditioning Factors (LCFs) sourced from different remote sensing techniques and the information of 197 historical landslide events acquired from the Geological Survey of India (GSI). The 5-fold stratified cross validation method was applied to generate training and testing dataset in different iterations. Four different tree-based ML classifiers including Decision Trees (DT), Random Forest (RF), Extreme Gradient Boosting (XGBoost) and Categorical Boosting (CatBoost) were employed to prepare the models to predict the landslide prone areas of the district. These classifiers were specifically chosen because they have the capability to handle feature importance and do not require separate feature selection methods, which are often subjective and difficult to standardize. These ensemble models were then evaluated using different performance metrics that are generally used to evaluate classification models. CatBoost classifier exhibited superior performance, achieving an accuracy of 87.93%, with a precision of 0.85, recall of 0.913, F1-score of 0.88, and an AUC-ROC value of 0.95. Although the RF model also demonstrated strong and competitive performance across all the evaluation metrics, CatBoost was selected for the final preparation of landslide susceptibility map (LSM) due to its comparatively higher recall and AUC-ROC values, which are critical for reliably identifying landslide-prone areas. Consequently, the final LSM was generated using the CatBoost model. According to the LSM, approximately 20.53% of the total district falls within the range of high susceptibility with prediction probability values ranging from 0.6 to 1.0. A major portion, around 47.04% lies within moderate susceptibility zones, and the remaining percentage corresponds to places that are relatively safe from slope failures. Furthermore, the feature importance scores extracted from CatBoost model revealed that slope, rainfall, soil type and distance to road are the main factors that contribute to triggering slope failures in the study area. The application of reverse geocoding techniques on the final LSM indicated that, southwestern and southern taluks including Mudigere, Sringeri, border regions of Koppa and the southern parts of Chikmagalur exhibit a high concentration of landslide prone areas compared to other places. This map serves as a critical tool for early warning systems and informed decision-making to reduce landslide risks in the district.

By carefully changing images to confuse deep learning models, adversarial attacks pose a serious risk to image recognition systems. Traditional defense methods often depend on models specific to each attack. This reliance raises costs and lacks the flexibility needed to tackle new threats, especially against emerging “invisible” attacks. A strong and scalable solution uses a single trained model based on the Pix2Pix GAN (Generative Adversarial Network) for image-to-image translation. This defensive strategy rebuilds the altered images before classification, sharply increasing classification accuracy from almost zero against attacks to an average of 86%. The model displays impressive adaptability by effectively countering various known and unknown attacks, such as FGSM, PGD, and BIM. Its reliable performance across different attack intensities shows its effectiveness. This image restoration method is a more efficient and scalable option compared to traditional adversarial training.

The energy efficiency of heat exchangers can be enhanced by integrating passive methods, such as tubular fin turbulators, nanofluids, and helical coil tubes. This study experimentally investigates the synergistic effects of tubular fin turbulators in conjunction with two types of nanofluids: graphene oxide (GO) and alumina (Al2O3) focusing on the heat transfer characteristics of a helically wrapped coil-in-shell heat exchanger (CSHE). The experiment uses three distinct helically wound coils and is conducted at a constant heat flux of 4 kW m−2. Two of the coils are equipped with tubular fins brazed to their outermost annular surfaces at orientations of α = 45° and 90°, while the third coil is a plain design without fins. The GO and Al2O3 nanofluids are used at volume concentrations of 0.05, 0.10, and 0.15

Product ordering workflows in small and medium-sized businesses (SMEs) conventionally depend on person-to-person methods like phone calls, physical handwritten documentation, and verbal acknowledgements. This manual approach frequently results in miscommunication, mistakes in orders, and considerable operational inefficiencies. This article introduces an AI-Powered Voice-Enabled Real-Time Product Ordering System, engineered to simplify and automatically manage the ordering interactions between shopkeepers and suppliers (stockholders). The architecture incorporates sophisticated speech recognition, natural language processing (NLP), and automated invoicing capabilities to facilitate precise order submission via spoken commands. The suggested methodology is built upon a five-part modular design: modules for voice interaction, understanding natural language, order handling, invoicing and notifications, and user permission control. The system's technical implementation relies on a Python Flask server, specialized speech recognition Application Programming Interfaces (APIs), and a MongoDB data store, secured by role-based authentication. Initial evaluations confirm effective conversion of speech to text, reliable interpretation of voice instructions, and smooth operational flow across all system components. Furthermore, the system is equipped to offer smart stock recommendations based on past sales records, perform automatic calculations for GST and discounts, and issue immediate order confirmations through email and SMS. The final data shows substantial gains in operational speed, precision, and ease of use compared to older, manual ordering systems, proving its high utility for small and medium enterprise settings.

Cybersecurity relies heavily on intrusion detection systems (IDSs) to secure business communication, as they serve as the primary line of defense against malicious activities. Existing research has highlighted the limitations of applying deep learning (DL) and machine learning (ML) approaches broadly to intrusion detection. However, the evaluation of the traditional model is often hindered by imbalanced datasets, limiting an accurate evaluation of model efficacy. A novel hybrid deep learning-based network intrusion detection method utilizing enhanced Bidirectional Long Short-Term Memory (Bi-LSTM) and Gated Recurrent Unit (GRU) approaches is introduced in this study to address these difficulties. Pre-processing, data augmentation, feature selection, and detection comprise the four stages of the paper. Initially, Min-Max normalization and null value handling are used to pre-process network traffic data. After pre-processing, the Improved Synthetic Minority Oversampling Technique (ISMoT) is employed to mitigate the issue of class imbalance. Then, features are chosen using the chaotic bat Artificial Bee Colony optimization method (CBABCOA). Non-negative matrix factorization (MNF) is employed for latent feature transformation to avoid overfitting problems during training. Furthermore, the NMF features are input to the Hybrid Residual one-dimensional convolutional bidirectional LSTM transformer (Hy-BiLSTM-Tr) and the auto-encoder-gated recurrent unit (AE-GRU) individually for performance comparison. The experimental results of the proposed model are evaluated using five datasets: UNSW-NB15, CICIDS 2019, NSL-KDD, CICIDS 2017 and BoT-IoT. The results show that Hy-BiLSTM-Tr and AE-GRU achieve high accuracies of 99.47 % and 98.42 %, respectively, which is strongly influenced by the quantity and quality of the dataset for intrusion detection.