As a predominant renewable energy source, solar energy has gained a significant prominence due to its sustainable characteristics and environmental benefits. Precise forecasting of photovoltaic power generation constitutes a critical requirement for optimal energy management and the maintenance of stability in electrical grid operations. This work proposes a novel predictive modeling framework that implements long-short-term memory (LSTM) neural networks through the KNIME Analytics Platform to predict solar energy production utilizing high-resolution sensor data. The comprehensive data set incorporates time series measurements of key solar parameters, including solar irradiance, global horizontal irradiance (GHI) power density, plane-of-array (POA) energy yield, and module temperature, acquired from an industrialgrade solar monitoring infrastructure. Data preprocessing techniques and visualizations were used to improve model performance and interpretability. The LSTM model effectively captured temporal dependencies in the data, with evaluation metrics including mean absolute error (MAE), mean squared error (MSE), and root mean squared error (RMSE). The implemented LSTM architecture successfully learned complex temporal patterns inherent in the solar generation data, with rigorous performance validation conducted using standard evaluation metrics: mean absolute error (MAE), mean squared error (MSE) and root mean squared error (RMSE). The results demonstrate that a well-optimized LSTM can effectively model long-term solar patterns without the computational overhead of hybrid architectures while maintaining competitive accuracy, making it more scalable and deployable for real-world solar energy forecasting. This contribution addresses a fundamental research gap in the field by providing a simplified yet highly accurate forecasting solution, particularly beneficial for grid operators and renewable energy planners seeking efficient and reliable predictions.

Modern Retrieval Augmented Generation often lacks an inherent understanding of document-specific relationships and structured knowledge. By combining large language models and graph-based retrieval, the PDF Query Chatbot presented in this research fills this gap and provides more precise and contextually aware responses. In order to explicitly record entity relationships and structural dependencies, the system uses Neo4j to create a knowledge graph after extracting textual content from the uploaded documents. To facilitate a semantic similarity search, the text is simultaneously shredded and embedded in a vector store. A hybrid retrieval system that combines vector-based search for contextual relevance and graph traversal for relational comprehension is activated when a user submits a query. To produce grounded, document-specific responses, the results of the two retrieval pipelines were combined and sent to the LLM. By synergizing graph databases, semantic search, and LLMs, this architecture provides a context-aware solution for intelligent document interaction, addressing key limitations in traditional LLM-based question resolution systems.

In today's digital world, processing multilingual documents is critical for business, legal tasks, and information retrieval. This study describes a Multilingual Document Processing System that uses Optical Character Recognition (OCR) and Retrieval-Augmented Generation (RAG) to extract, query and summarize text in multiple languages. The system employs advanced OCR models to correctly recognize text from scanned documents, images, and handwriting in various scripts. By incorporating RAG, it improves comprehension and response generation, allowing users to retrieve and summarize information in English even when the original language is different. This approach takes advantage of recent advances in natural language processing, large language models (LLM), and multimodal AI to address challenges in multilingual data accessibility, knowledge synthesis, and real-time communication. The system provides a scalable AI-driven solution to improve document processing, eliminate language barriers, and increase global user engagement. AWS services support scalable document processing but cold starts in AWS Lambda hinder real time tasks.

There is an increasing need for intelligent systems to improve interior design as smart home technologies advance. In this paper, we present a new method for recommending room decor, which makes use of the object identification model YOLOv3 (You Only Look Once version 3). In contrast to the conventional techniques, the proposed model makes decor recommendations for a room based on the necessary furniture and dimensions of the room. The proposed model is able to precisely identify and categorize different types of furniture and decor inside a picture of a room. Users receive customized recommendations that show them the way the decor items selected by them look in their living area and the size of items that suit that area. The degree of confidence with which the model can identify and suggest an image from the data set including the user-selected items is used to validate the results. The proposed model is tested with the products that are connected to the free IKEA APIs, which provided suggestions for every iteration of the item chosen by the user.

Accurate product classification in e-Commerce and supply chain management is essential to smooth operations and enhance the customer experience. While Large Language Models (LLMs) perform exceptionally in natural language processing, they encounter issues like model hallucination and dependence on outdated information. Furthermore, LLMs often rely on outdated data. This paper introduces an open source cloud-based RAG model, using Amazon Web Services (AWS) and vector databases to address these issues. The RAG architecture combines retrieval-based and generation-based methods, allowing them to supplement responses with up-to-date information from external sources, thus reducing the risk of model hallucination. The project employs a Vector DB deployed in EC2 to improve contextual understanding and retrieval capabilities of these large language models. Through comprehensive experimentation and AWS deployment, the RAG system improved contextual comprehension and increased the accuracy of the generated output. Semantic similarity search results significantly improve retrieval performance.

Indian courts have too many cases and not enough resources or judges, causing delays and undermining trust in the legal system. In addition, Indian law is quite complex, requiring law students and legal professionals to study and understand it thoroughly for better practice. To address these challenges, we have developed an AI-powered legal assistant designed to enhance the efficiency and effectiveness of legal processes. The three main features of our proposed approach are judgment prediction, case summarization, and IPC section and punishment prediction. The judgment prediction uses AI & ML to analyze case facts and give the probability of each party winning. The case summarization compacts the long legal documents into concise texts, highlighting key legal terms and facts. The IPC section and punishment prediction identifies applicable sections of the Indian Penal Code and predicts. These features enhance legal workflows, reduce manual document analysis and improve legal decision-making. Our proposed legal AI system is tested by deploying on Streamlite, which provides interactive interface that can be used on any device with internet access. The performance of the proposed AI model is analysed and the general accuracy of the classification model is obtained as 97%. The performance can be further improved using a more comprehensive and diverse dataset. The consistency and reliability of the model are validated by computing macro- and weighted averages, affirming its robust performance across all IPC categories. In addition, the precision, recall, and F1 score values indicates the accurate identification of winning parties.

Power consumption forecasting plays a critical role in effective energy management and resource allocation. However, accurate prediction of power consumption presents unique challenges due to the diverse nature of energy use patterns. This intricate interplay requires the application of advanced forecasting methodologies capable of effectively capturing the temporal dynamics and inherent non-linear relationships embedded within power consumption data across diverse time horizons. In accordance with these findings, this research delves into the complexities of the prediction of energy consumption by proposing a novel approach. We leverage long-short-term memory (LSTM) networks, a type of recurrent neural network (RNN), for long-term energy usage predictions due to their ability to model temporal dependencies. We utilized a polyregression model for short-term forecasting tasks, capitalizing on its effectiveness with smaller, non-cyclical datasets. The results demonstrate the superiority of LSTMs in handling complex relationships within power usage data. In contrast, the polyregression model, while achieving an acceptable R2 of 0.763 for smaller datasets, struggled with limited data points. These findings contribute significantly to the field of power usage forecasting by highlighting the effectiveness of long-Short-Term Memory networks in handling large, featurerich datasets and providing a robust and reliable approach for utilities and energy providers seeking to improve their load forecasting capabilities, leading to more efficient resource management and grid operation.

Assessment of a person's mental health is a complex phenomenon that affects many people around the world. This study aims to develop an accurate and efficient method to identify possible mental health concerns through text analysis, such as written messages and social media posts. This paper uses the capabilities of Amazon SageMaker Autopilot and KNIME to deliver improved model performance. This approach yields an improvement in accuracy compared to baseline models, demonstrating the potential to take advantage of cloud-based machine learning platforms and data science tools to streamline workflows. The results of the experiment demonstrate the effectiveness of the proposed approach. Specifically, XGBoost achieved the highest accuracy of 93.6%, outperforminag Linear Learner and Multilayer Perceptron.

This work explores the critical role of autoverification in laboratory medicine, where timely and accurate test results are imperative for effective patient care. Auto-verification systems can significantly reduce the time to report and enhance the reliability of test outcomes, which is particularly crucial in time-sensitive diagnostic environments. This study delves into meticulous preprocessing of clinical data to prepare them for analysis, addressing challenges such as data inconsistency and missing values. By integrating advanced machine learning (ML) and deep learning (DL) models, we develop robust algorithms aimed at automating the verification of laboratory test results. Furthermore, we demonstrate the feasibility of these algorithms by replicating the processes in KNIME, a data analytics platform. This not only substantiates the scalability of our approach but also underscores its potential for real-world application in improving diagnostic workflows and patient outcomes.

The Internet of Things has spurred urban growth and increased environmental pollution. This paper introduces EcoCoin, a mobile app that rewards college campus students for sustainable actions. In addition, it explores how facial recognition can deliver personalized ads based on expressions, improving the retail experience with real-time recommendations. The paper concludes with an overview of Amazon cloud services' advanced features, underscoring their utility for developers to incorporate sophisticated image and video analysis, and highlighting technology's role in achieving social goals.

This work proposes a cloud-based attendance system that uses face recognition technology to authorize identity. The system uses the Amazon Web Services (AWS) Rekognition service and a serverless architecture. The proposed system provides a reliable and tamperproof solution to track attendance, eliminating the need for manual record keeping and minimizing human involvement. It also offers potential benefits, such as improved security and transparency in attendance management.

The digital revolution has made access to information very easy. The onset of the COVID-19 pandemic also called for further digitization. Every organization; be it an office, an educational institute or a government entity, was forced to shift to an all virtual mode of operation. This led to the conduct of online examinations with very little time for formulating an anti-cheating examination pattern. Audio and video proctoring tools are considered helpful but are very expensive and do not provide a method to detect plagiarism in the handwritten text. This is a serious problem for academic enterprises and institutes where there is a need for plagiarism detection in the submitted assignments, answer-scripts against the information available on the Internet as well as against other submissions. This paper presents a plagiarism detection system for handwritten text in English. The proposed system uses authentication tools/services, cloud storage, and optical character recognition (OCR) services to automate the process of checking plagiarism between two handwritten documents, as well as plagiarism with respect to all information available online.

The advancements in Computer Vision with Deep Learning are playing an essential role in aiding Healthcare Organizations to offer better patient care while reducing costs and improving efficiencies. The medical industry has seen a sharp growth since the past few decades and it has benefited thousands of living beings. Skin being the largest part of human beings, it has never been given the proper care nor importance, hence widely resulting in the ignorance of skin infections or cancers. Despite the diagnosis of several such diseases being detected at the later or critical stages, there is a little to no scope of detection where the affected people can identify the symptoms immediately and could reach a physician. In this paper, we present the development of a model using deep learning approach to identify the specific dermatitis and a mobile application that functions in a way to detect the disease when a picture of the symptomatic part is taken and uploaded. This scheme assists the medical professionals and patients during pandamic situations like COVID 19. This application is user friendly and very much useful for the people living in rural areas and hilly area where it takes long time and expenditure to reach hospitals.

Wake-up stroke refers to a kind of ischemic stroke where a person wakes up with symptoms of stroke that are not present before going to sleep. These symptoms may include muscle weakness, drowsiness, walking difficulties, face drooping among others. Risk factors for ischemic strokes are Diabetes, Hypertension, Obesity, age, tobacco use, etc. From the Statistical Findings, 8-28 percent of all brain ischemic strokes consist of Wake-up strokes. The main method of treating an ischemic stroke is tissue Plasminogen Activator (tPA), the use of which is approved for 3-4.5 hours from the onset time of stroke. As the onset time in wake-up strokes is difficult to determine, the person may not be eligible for tPA treatment. This paper presents the development of a prototype to detect the wake-up stroke using the appropriate sensors to obtain physiological data and Internet of things technology to issue an alert signal to the concerned people of the patient. In India, various factors causes the delay of the patient's arrival to the hospital. The proposed device helps the people to take the patient to the hospital within time to reduce the risk of permanent disabilities like paralysis and memory loss.

Due to medical advancements in 20th century there has been an explosive growth in the population and life expectancy. Even after such medical advancements there still exist some irreversible diseases from which modern medicine cannot help humans to recover fully. Alzheimer's is one such progressive disease that causes severe problems with thinking, memory and behaviour. The people suffering from Alzheimer's disease have a tendency to forget people they know, scheduled meetings, medicines they are supposed to take, their daily routine and other important things. In this paper, we present the prototype developed for wearable smart cap designed using deep learning model for face recognition to recognize people related to the patient. In contrast to the other approaches which rely on deep networks trained on specific number of classes making introduction of new classes difficult in the system, our approach extracts unique feature vectors from faces in the dataset and trains classifier on them so that it can easily be trained for classifying new classes. We have developed a 3D model for this smart cap with all hardware modules embedded in it. This cap is very much useful for the patients as it enables them to recognize the people nurturing them and also harmful strangers.

Air pollution is one of the vital issues to ponder in current environmental situation as it has major impact on human health and environment. Real time monitoring of air pollution will help in calculating air quality index to issue health advisories as well as for taking necessary actions to meet standards. Air pollutants in the form of ground-level ozone and particulate matter have been the major pollutants in recent times. The proposed prototype in this paper deals with a smart Mongoose OS based monitoring system to monitor the harmful pollutants' concentration continuously. Mongoose OS is a cross-platform IoT operating system, which provides generic infrastructure layer for smart products and minimal foot print on edge device. As Monitoring is carried out uninterrupted, concerned officials get notified whenever a certain pollutant surpasses the threshold. Prototype is designed using cost effective low power ESP32 development board and appropriate sensors to monitor CO, CO2, NH3, Smoke particulates ( PM2.5) and Ozone. The values of pollutants (unit of measurement is PPM) captured by this prototype can be sent through message brokers to the cloud server set up using Losant IoT platform, by configuring it as Wi-Fi access point. PPM values of pollutants are then displayed on to losant device logs through MQTT protocol. Losant device ID, API access tokens play a key role in providing access to the users on Losant.

The concentration of pollutants in the environment is increasing at a rapid rate and one of the major sources of air pollution is vehicular emission. Vehicle emissions mainly constitute of Carbon monoxide, Nitrogen oxide and other toxic gases. These gases when present in the lower atmosphere adversely affect the health of human beings and causes respiratory problems like asthma, stroke, emphysema, lung cancer etc. Since this issue needs an immediate attention, environmentalist and other government bodies are coming up with new techniques. Health of the people can be improved by monitoring the air pollution level and notifying them whenever the quality of air degrades beyond a certain level. This paper proposes a system which incorporates the Radio Frequency Identification (RFID) technology for detecting the vehicles emission level and notifying the vehicle owner and the concerned authorities if the measured values exceed the standard limits for taking appropriate action. The proposed system also employs Maximum Spanning Tree(MAXST) algorithm to optimize the number of readers to be installed and hence reduces the installation cost.

Having one or the other form of security system is a must as it acts as the first line of defence in case of any break-ins. Houses with no advanced security systems usually have a higher chance of being targeted than those which are installed with sophisticated security systems. The essential part of security is intruder detection system. Till date several solutions have been proposed which make use of PIR sensor and in most of them, the owner is notified every time the sensor detects a motion. This leads to several false alarms as it might not always be an intruder. The solution proposed in this paper aims to reduce this false alarm rate. The proposed solution leverage's the human tendency to carry their mobile phones with them wherever ever they go and their habit to use the Internet services while being connected to the home access point rather than mobile data while at home. The security system uses a PIR sensor and is imparted with contextual/environmental awareness, which will let it take better decision as to when to notify the owner and hence reducing the false alarm rate. The contextual awareness of the system is possible due to the promiscuous mode of operation in ESP8266. The system scans for Wi-Fi packets and identifies the origin and destination MAC addresses of the devices communicating. The system then checks for the MAC address of the owner's phone to know his/her presence. Then it decides whether or not to notify the owner in case of any motion. The Instance of intrusion will also be logged in a SQL database.

The performance of many data security and reliability applications depends on computations in finite fields GF (2m). In finite field arithmetic, field multiplication is a complex operation and is also used in other operations such as inversion and exponentiation. By considering the application domain needs, a variety of efficient algorithms and architectures are proposed in the literature for field GF (2m) multiplier. With the rapid emergence of Internet of Things (IoT) and Wireless Sensor Networks (WSN), many resource-constrained devices such as IoT edge devices and WSN end nodes came into existence. The data bus width of these constrained devices is typically smaller. Digit-level architectures which can make use of the full data bus are suitable for these devices. In this paper, we propose a new fully digit-serial polynomial basis finite field GF (2m) multiplier where both the operands enter the architecture concurrently at digit-level. Though there are many digit-level multipliers available for polynomial basis multiplication in the literature, it is for the first time to propose a fully digit-serial polynomial basis multiplier. The proposed multiplication scheme is based on the multiplication scheme presented in the literature for a redundant basis multiplication. The proposed polynomial basis multiplication results in a high-throughput architecture. This multiplier is applicable for a class of trinomials, and this class of irreducible polynomials is highly desirable for IoT edge devices since it allows the least area and time complexities. The proposed multiplier achieves better throughput when compared with previous digit-level architectures.

Orthogonal frequency division multiplexing (OFDM) is an emerging technology in recent wireless communication standards where high data rate is required at low latency and better spectral efficiency. OFDM signals have a generic problem of high peak‐to‐average power ratio (PAPR) due to the superimposition of the data‐modulated subcarriers. This high peak signals may result in amplifier non‐linearity which creates many problems such as performance degradation and out‐of‐band distortion. Hence, the authors addressed this problem by designing and implementing four new PAPR reduction schemes such as phase modulation, rail clipping, sample and hold approach and threshold methods based on amplitude clipping to reduce PAPR and improve the band gap between spectral sidelobes to main lobe significantly. For practical proof of the suggested concepts, they have chosen software‐defined radio as an experimental setup in which universal software radio peripheral N210 employed as hardware and GNU radio as software. Experimental results are analysed in terms of significant PAPR reduction and out‐of‐band spectral leakage control. Besides these, motivation, background and relevant mathematical analysis with rigid justification are presented in this study.

The high peaks of a conventional orthogonal frequency division multiplexing (OFDM) symbol may push the amplifier into the non‐linear region, which creates many problems that reduce performance and lead to out of band distortion. In this study, the authors present and analyse four different peak to average power ratio (PAPR) reduction techniques such as phase modulation, rail clipping, sample and hold approach and threshold method for packet data transmission of OFDM on mobile‐worldwide interoperability for microwave access system. The performance of each PAPR reduction technique on OFDM system is presented in time and frequency domains separately. Therefore an existing OFDM system can provide an additional CE‐OFDM mode with relative ease, particularly for the case of software defined platforms. Nowadays, the behaviour of a communication system has been modified by simply changing its software. This gave rise to a new radio model called software defined radio, in which all hardware components are implemented in software rather than in hardware. For our investigation, GNU radio and universal software radio peripheral N210 are employed as software and hardware platforms, respectively. It is shown that for a high clipping threshold value, the gap between in‐band and out‐of‐band radiations is maintained well.