St. Francis Institute of Management and Research

Foreign Portfolio Investment (FPI) plays a vital role in deepening financial markets, enhancing liquidity, and integrating emerging economies like India into the global capital system. Unlike Foreign Direct Investment (FDI), FPI involves short-term, non-controlling investments in equities, bonds, and hybrid instruments, making it sensitive to global economic, financial, and political conditions. This study examines the determinants of Net FPI inflows into India across pre-COVID (April 2012–March 2020) and post-COVID (April 2021–March 2025) periods, highlighting shifts in investor behavior and market dynamics. Using monthly data from RBI, NSDL, NSE, MOSPI, Bloomberg, and S&P Dow Jones, the study applies the Augmented Dickey-Fuller (ADF) test for stationarity and the Autoregressive Distributed Lag (ARDL) model to analyze short-run relationships between Net FPI and key domestic and global factors, including inflation (CPI), industrial production (IIP), broad money supply (M3), stock market performance (Nifty 50 and S&P 500 growth), exchange rate (USD/INR), market volatility (NSE VIX and USA VIX), and the Global Economic Policy Uncertainty (GEPU) index. Results indicate that pre-COVID inflows were influenced by past FPI trends, domestic inflation, exchange rate movements, stock market returns, and global policy uncertainty, while industrial production and S&P 500 returns had counter-cyclical effects. Post-COVID, determinants shifted toward domestic growth fundamentals, currency movements, and global uncertainty, with inflation and volatility losing significance. Global market performance, particularly the S&P 500, consistently showed a substitution effect. Diagnostic tests confirm model stability, and R-squared values indicate strong explanatory power. The study suggests that post-COVID, foreign investors prioritize growth fundamentals and global risk over short-term domestic fluctuations. For policymakers, this underscores the importance of macroeconomic stability, clear communication, and effective risk management to attract and sustain FPI. By integrating long-term monthly data with advanced econometric techniques, the research provides actionable insights for portfolio managers, regulators, and corporate strategists navigating India’s dynamic investment landscape.

Leg bone fractures require accurate and timely diagnosis to determine appropriate treatment. This research proposes an Artificial Intelligence (AI) based deep learning framework for automated leg bone fracture detection and severity assessment using X-ray images. Initially, a Median Filter is applied as a preprocessing technique to remove noise and enhance fracture-related structures. The preprocessed images are then analyzed using Convolutional Neural Network (CNN) and Artificial Neural Network (ANN) algorithms for feature extraction, fracture detection, and classification. The proposed system identifies fractured and non-fractured bones and assesses the severity or stage of the fracture. Based on the detected fracture characteristics, the system provides computer-aided treatment recommendations, including conservative management, supportive medication, or surgical intervention when clinically appropriate. The proposed framework aims to improve fracture detection accuracy, reduce diagnostic time, and support healthcare professionals in making effective treatment decisions.

Decision-making on the selection-based problem of the location of Roadside units (RSU) in vehicular networks is complicated. This intricate problem shall be resolved with the intervention of Explainable Artificial Intelligence (XAI) and fuzzy-based multi-criteria decision-making methods (MCDM). This research proposes the integrated decision-making approach in making optimal location selection considering different criteria and economic aspects. The decision-making problem discussed in this chapter considers location selection of roadside units as the choice-making of the location is very significant in enhancing vehicular efficiency. This decision problem considers the criteria of coverage area, traffic density, accident frequency, infrastructure availability, environmental impact, energy efficiency, and cost components such as cost effectiveness, revenue competence, and investment returns. The integrated approach is leveraged to make an optimal choice of the RSU locations.

Cancer is one of the very dangerous diseases which cause lots of death all over the world. There are various types of cancers which affect male, female even children too. Among various types of cancer, the cervical cancer is one types which is found in women and which is a major cause of death in women, it is a very complicated disease but if it is detected at an early stage then we can limit its complications. It is possible to detect cervical cancer at an early stage with the help of various ways and means. There are categories of symptom-based, diagnosis-based, and many other ways the cervical cancer can be detected among women. Here in this research paper shows the use of ResNet50 and MobileNetv2 for diagnosis or detection of cervical cancer using image processing and deep learning approach. Here, after applying both the deep learning-based approaches, their comparative analysis is also performed to determine their efficiency in delivering the result in term of diagnosis the cervical cancer. To conduct this study cell images were fed into the CNN models, and they were trained to evaluate which model provides more accuracy than the other, five classes of images that were created are as follows Dyskeratotic, Koilocytotic, Metaplastic, Parabasal, Superficial-Intermediate, and the models were trained based on these five classes. After various preprocessing and augmentation methods that were conducted on the images, the models were trained, and the results showed that MobileNet-V2 was able to achieve 87

Breast cancer is a major global health concern and one of the leading causes of death among women. Its risk is influenced by both genetic and environmental factors, with diet recognized as a key modifiable contributor. This study focuses on the role of vegetarian and non-vegetarian food habits in influencing breast cancer progression and applies deep learning models to predict risk outcomes. Objective: The primary objective is to analyze how different dietary patterns, particularly plant-based versus animal-based diets, affect breast cancer risk and progression. Methods: A dataset of breast cancer patients was classified into vegetarian and non-vegetarian groups. Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) models were employed to predict patient risk levels and identify hidden patterns linking dietary intake with cancer growth. Findings: The results reveal that non-vegetarian diets, especially those high in fats, red meat, and processed foods, are associated with increased cell growth and higher cancer risk. In contrast, vegetarian diets rich in fiber, antioxidants, and phytochemicals demonstrate a protective effect. CNN achieved the highest accuracy, while LSTM effectively captured sequential dietary behaviors. These findings confirm that deep learning can serve as a valuable tool for diet-based risk prediction in breast cancer patients.