The Indian economy is strongly dependent on agriculture, with Karnataka playing a significant role in food production. Accurate crop yield forecasting is essential for data-driven decision making, efficient resource utilization, and sustainable agricultural management. However, traditional statistical and standalone machine learning models often fail to capture the complex, nonlinear, and time-varying interactions among crop characteristics, soil properties, and climatic factors. To address this limitation, this paper proposes SmartAgri-Net, a hybrid intelligent ensemble framework that integrates Decision Tree (DT) for interpretability, Gated Recurrent Unit (GRU) for sequential learning, and Long Short-Term Memory (LSTM) for temporal dependency modeling. The system utilizes the Raitamitra dataset provided by the Government of Karnataka, covering major crops such as rice, maize, ragi, and groundnut with agro-climatic and soil attributes. The methodology includes structured preprocessing, normalization, exploratory data analysis, feature selection, and weighted ensemble prediction. Experimental results demonstrate improved forecasting performance over individual models using MAE, RMSE, R2, and MAPE metrics. The proposed framework supports scalable, interpretable, and intelligent decision support for precision agriculture and policy planning, with future extensions toward real-time IoT-enabled systems.
Lung cancer is a deadly cancer in the world and therefore it is very much important that early and precise detection is done. PulmoXLM-Net is a two step lung cancer detector and classifier research that utilizes Xception, Liquid State Machines (LSMs) and Multi-Class SVM (MSVM) in this paper. In the first stage, preprocessing of CT images in the LIDC-IDRI dataset is done through median filtering and morphological operations and active contour segmentation. Refinement of tumor regions is performed by binary conversion, circle detection and area estimation. Xception network finds out the spatial and texture features of the segmented regions at the second stage that are dynamically added by LSMs to detect sequential patterns then classified into Adenocarcinoma, Large Cell Carcinoma, Squamous Cell Carcinoma and Normal through MSVM. The performance is determined with the help of accuracy, sensitivity, specificity, and precision that would demonstrate the effective segmentation and proper classification. PulmoXLM-Net is a decision support system that is good in lung cancer screening.
Machine learning techniques to predict student performance using a dataset of demographic, behavioral, and academic indicators are being used extensively in recent research. Using algorithms like support vector machines, random forests, and neural networks, the random forest algorithm achieves the highest predictive accuracy. The study provides insights into factors influencing student performance, such as engagement and academic achievement. It offers practical implications for educators and policymakers, enabling targeted interventions to improve learning outcomes.
This paper shows an IoT-based transformer safety system designed to protect humans and distribution equipment in real time. Using the ESP8266 NodeMCU, the system continuously monitors temperature, current, and human proximity using low-cost sensors. When a person or object approaches the transformer beyond a defined danger threshold, the controller disconnects the relay, activates an inductive discharge mechanism, and sends real-time alerts through Firebase and a Telegram bot deployed on PythonAnywhere. The proposed method prioritizes human safety, reduces accident risks near transformer yards, and enables remote monitoring and maintenance control. Unlike conventional monitoring systems, the proposed framework integrates real-time human-in-yard detection with an automated inductive discharge mechanism, ensuring immediate isolation and energy dissipation when a hazard is detected, thereby adding a proactive human-safety layer.
Customization of Emojis from real-time facial pictures is an alarming trend in personalization of applications. The real-time facial pictures captured containing salient facial landmarks are the candidates for personalized emojis. There are many ways to display emoji language cues. As personalized emojis have become popular forms of communication, existing solutions are in dearth of customization options. A new approach that uses facial landmarks captured by Python and combines them with a facial rig in Unity to create dynamic emojis is proposed. Our approach includes: (1) using Python libraries such as OpenCV or Media Pipe, capture facial landmarks, (2) exporting landmark information to Unity, and (3) dynamically creating a predefined front rig in Unity based on the locations of the received landmarks. The proposed framework allows users to create real-time facial expression emojis, increasing the emotional implications of digital communication. Our findings highlight the potential of this approach in terms of emojis visibility and the ability to customize interactive media. This research helps automate the human-computer interaction by providing a creative framework for creating dynamic emojis for the user.
