Home Menu
icon-menu-close.svg

Faculty & Research

Faculty & Research

Research Productive

Show result

Clear
Search Query :
Area :
Search Query :
849 items in total found

Journal Articles | 2024

Bayesian predictive inference for nonprobability samples with spatial poststratification

Dhiman Bhadra, Balgobin Nandram

Non-probability sampling involves selecting samples from a population in which the probability of selection is unknown and some population units may have zero selection probabilities. This differentiates it from probability sampling where selection is governed by a probability model and every population unit has a non-zero chance of being selected. Nonprobability samples usually suffer from selection bias and hence may not represent the target population accurately. An important problem that arises in this context is the prediction of responses corresponding to non-sampled units, which should ideally have been sampled. In this article, we propose three modeling frameworks to address this issue. We use propensity scores to balance the sampled and non-sampled units and a Bayesian estimation scheme for parameter inference and prediction. We incorporate a spatial poststratification scheme to assess the predictive ability of our models on a simulated dataset. In addition, we perform model selection routines to identify the optimal model having the best predictive ability.

Go to link
Read More

Journal Articles | 2024

Interpretable classifier models for decision support using high utility gain patterns

Srikumar Krishnamoorthy

Ensemble models such as gradient boosting and random forests are proven to offer the best predictive performance on a wide variety of supervised learning problems. The high performance of these black box models, however, comes at a cost of model interpretability. They are also inadequate to meet regulatory demands and explainability needs of organizations. The model interpretability in high performance black-box models is achieved with the help of post-hoc explainable models such as Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP). This paper presents an alternate intrinsic classifier model that extracts a class of higher order patterns and embeds them into an interpretable learning model. More specifically, the proposed model extracts novel High Utility Gain (HUG) patterns that capture higher order interactions, transforms the model input data into a new space, and applies interpretable classifier methods on the transformed space. We conduct rigorous experiments on forty benchmark binary and multi-class classification datasets to evaluate the proposed model against the state-of-the-art ensemble and interpretable classifier models. The proposed model was comprehensively assessed on three key dimensions: 1) quality of predictions using classifier measures such as accuracy, F1 , AUC, H-measure, and logistic loss, 2) computational performance on large and high-dimensional data, and 3) interpretability aspects. The HUG-based learning model was found to deliver performance comparable to that of the state-of-the-art ensemble models. Our model was also found to achieve 2-40% (45%) prediction quality (interpretability) improvements with significantly lower computational requirements over other interpretable classifier models. Furthermore, we present case studies in finance and healthcare domains and generate one- and two-dimensional HUG profiles to illustrate the interpretability aspects of our HUG models. The proposed solution offers an alternate approach to build high performance and transparent machine learning classifier models. We hope that our ML solution help organizations meet their growing regulatory and explainability needs.

Go to link
Read More

Journal Articles | 2024

AI as an inventor debate under the Patent Law: A post-DABUS comparative analysis

Sarvanan A Deva, Prasad M

Artificial Intelligence (AI) has gained momentum during the last decade, achieving narrow intelligence to general intelligence. The Gen AIs can generate original content and create patentable inventions. The key challenge is whether AI machines can be considered as "inventors" under the existing patent laws; this question has been challenged before various domestic courts. In this context, this paper analyses the interplay between AI and existing patent law frameworks; more specifically, we look at a post-DABUS case from a comparative perspective. The paper explores this question predominantly from the viewpoint of Australia, the United Kingdom, and the United States. The limitations of the existing patent regime and the need to adopt a flexible one that could adapt to the challenges posed by AI are highlighted. The crucial factors shaping the future direction of patent law in the context of AI debate, such as the evolving need for a sui generis law and convergence for a global standard, also form a part of this paper.

Go to link
Read More

Journal Articles | 2024

𝛿- perturbation of bilevel optimization problems: An error bound analysis

Margarita Antoniou, Ankur Sinha, Gregor Papa

In this paper, we analyze a perturbed formulation of bilevel optimization problems, which we refer to as -perturbed formulation. The -perturbed formulation allows to handle the lower level optimization problem efficiently when there are multiple lower level optimal solutions. By using an appropriate perturbation strategy for the optimistic or pessimistic formulation, one can ensure that the optimization problem at the lower level contains only a single (approximate) optimal solution for any given decision at the upper level. The optimistic or the pessimistic bilevel optimal solution can then be efficiently searched for by algorithms that rely on solving the lower level optimization problem multiple times during the solution search procedure. The -perturbed formulation is arrived at by adding the upper level objective function to the lower level objective function after multiplying the upper level objective by a small positive/negative . We provide a proof that the -perturbed formulation is approximately equivalent to the original optimistic or pessimistic formulation and give an error bound for the approximation. We apply this scheme to a class of algorithms that attempts to solve optimistic and pessimistic variants of bilevel optimization problems by repeatedly solving the lower level optimization problem.

Go to link
Read More

Journal Articles | 2024

Ambient air pollution and daily mortality in ten cities of India: A causal modelling study

"Jeroen de Bont, Bhargav Krishna, Massimo Stafoggia, Tirthankar Banerjee, Hem Dholakia, Amit Garg, Vijendra Ingole, Suganthi Jaganathan, Itai Kloog, Kevin Lane, Rajesh Kumar Mall, Siddhartha Mandal, Amruta Nori-Sarma, Dorairaj Prabhakaran, Ajit Rajiva, Ab

Background

The evidence for acute effects of air pollution on mortality in India is scarce, despite the extreme concentrations of air pollution observed. This is the first multi-city study in India that examines the association between short-term exposure to PM2·5 and daily mortality using causal methods that highlight the importance of locally generated air pollution.

Methods

We applied a time-series analysis to ten cities in India between 2008 and 2019. We assessed city-wide daily PM2·5 concentrations using a novel hybrid nationwide spatiotemporal model and estimated city-specific effects of PM2·5 using a generalised additive Poisson regression model. City-specific results were then meta-analysed. We applied an instrumental variable causal approach (including planetary boundary layer height, wind speed, and atmospheric pressure) to evaluate the causal effect of locally generated air pollution on mortality. We obtained an integrated exposure–response curve through a multivariate meta-regression of the city-specific exposure–response curve and calculated the fraction of deaths attributable to air pollution concentrations exceeding the current WHO 24 h ambient PM2·5 guideline of 15 μg/m3. To explore the shape of the exposure–response curve at lower exposures, we further limited the analyses to days with concentrations lower than the current Indian standard (60 μg/m3).

Findings

We observed that a 10 μg/m3 increase in 2-day moving average of PM2·5 was associated with 1·4% (95% CI 0·7–2·2) higher daily mortality. In our causal instrumental variable analyses representing the effect of locally generated air pollution, we observed a stronger association with daily mortality (3·6% [2·1–5·0]) than our overall estimate. Our integrated exposure–response curve suggested steeper slopes at lower levels of exposure and an attenuation of the slope at high exposure levels. We observed two times higher risk of death per 10 μg/m3 increase when restricting our analyses to observations below the Indian air quality standard (2·7% [1·7–3·6]). Using the integrated exposure–response curve, we observed that 7·2% (4·2%–10·1%) of all daily deaths were attributed to PM2·5 concentrations higher than the WHO guidelines.

Interpretation

Short-term PM2·5 exposure was associated with a high risk of death in India, even at concentrations well below the current Indian PM2·5 standard. These associations were stronger for locally generated air pollutants quantified through causal modelling methods than conventional time-series analysis, further supporting a plausible causal link.

Funding

Swedish Research Council for Sustainable Development.

 

Go to link
Read More

Journal Articles | 2024

Demonetisation and labour force participation in India: The impact of governance and political alignment

"Pritha Dev, Jeemol Unni"

This paper considers the demonetisation event of November 2016 in India and its impact on labour force participation and incomes with a view to understand the role of governance in how individuals weathered the shock. We measure governance at the state level by measuring if the state was ruled by Bhartiya Janta Party at the time of demonetisation which was also the ruling party in the centre under whom demonetisation was introduced. We use a difference in difference approach to compare outcomes before and after demonetisation for individuals in politically aligned states. We control for financial inclusion by the banking penetration at the state level and whether individuals had a bank account. We find evidence of a gendered impact of the macroeconomic shock of demonetisation. While we find that demonetisation is associated with an overall drop in labour force participation, we find that individuals residing in BJP ruled states fared relatively better and this was particularly true for females. We find that while incomes continued to rise post demonetisation, individuals residing in BJP ruled states had relatively better outcomes. We additionally controlled for states which were up for election in 2017 and find that individuals in states which were BJP ruled and up for reelections showed better employment and income outcomes.

Go to link
Read More

Journal Articles | 2024

The strongest link: Service-profit chain as a conduit for enabling entrepreneurial orientation and multidimensional service performance

Gurjeet Kaur Sahi, Rupali Mahajan, Anand Kumar Jaiswal, Pankaj C. Patel

The paucity of literature to integrate the duality of customer and employee strategic orientations of an organization with its service operational inputs for achieving superior returns has provided an impetus to examine the significance of internal processes and behaviors as a mediating mechanism in the entrepreneurial orientation–firm performance relationship. Grounded in dynamic capabilities theory, in this article, we aim to extend the previous literature on operations management by investigating the role of the service-profit chain (SPC) in relating entrepreneurial orientation with performance—new products, subscribers, and revenues—in the telecom service industry. The study draws on archival performance data provided by the Telecom Regulatory Authority of India, combined with the primary data collected from 179 managers working in private telecom organizations in India. The findings of our study show that entrepreneurial orientation positively influences new product success only if internal and external SPCs are strong. However, entrepreneurial orientation is negatively and nonsignificantly related to gross revenue. Subscriber base, another important factor that helps in establishing the positive and significant link between entrepreneurial orientation and gross revenue, is higher under stronger internal SPC. Thus, the managers must emphasize more on employee job conditions to pursue strong customer orientation in improving service operations. The findings have implications for strategic and operations’ considerations in a service context.

Go to link
Read More

Journal Articles | 2024

Mitigation of non-CO2 greenhouse gases from Indian agriculture sector

Omkar Patange, Pallav Purohit, Vidhee Avashia, Zbigniew Klimont, Amit Garg

The Indian agriculture sector is driven by small and marginal farmers and employs two-thirds of the Indian work force. Agriculture also accounts for around a quarter of the total greenhouse gas emissions, mainly in the form of methane (CH4) and nitrous oxide (N2O). Hence, agriculture is an important sector for India’s transition to net-zero emissions and for the achievement of the sustainable development goals. So far, very few studies have assessed the future trajectories for CH4 and N2O emissions from the agriculture sector. Moreover, assessment of CH4 and N2O mitigation potential at a subnational (state) level is missing but is important owing to the regional diversity in India. To fill this gap, we focus on methane and nitrous oxide emissions from the agricultural activities using 23 sub-regions in India. We use the GAINS modelling framework which has been widely applied for assessing the mitigation strategies for non-CO2 emissions and multiple air pollutants at regional and global scales. We analyze a current policy and a sustainable agriculture scenario using different combinations of structural interventions and technological control measures to inform the Indian and global climate policy debates. Our results suggest that a combination of sustainable agricultural practices and maximum feasible control measures could reduce the CH4 and N2O emissions by about 6% and 19% by 2030 and 27% and 40% by 2050 when compared to the current policies scenario with limited technological interventions. At a sub-national level, highest mitigation potential is observed in Uttar Pradesh, followed by, Madhya Pradesh, Rajasthan, Gujarat, Maharashtra, Andhra Pradesh, and Telangana. The mitigation of agricultural CH4 and N2O also has co-benefits in terms of reduced local pollution, improved health, and livelihood opportunities for the local communities.

Go to link
Read More

Journal Articles | 2024

The capacitated r-hub interdiction problem with congestion: Models and solution approaches.

Sneha Dhyani Bhatt, Ankur Sinha, Sachin Jayaswal

We study the 𝑟-hub interdiction problem under the case of possible congestion. Hub interdiction problems are modeled as attacker-defender problems to identify a set of 𝑟 critical hubs from a set of 𝑝 hubs, which when attacked, causes maximum damage to network restoration activities of the defender. In this work we consider that in addition to the routing cost, the defender also aims to minimize the congestion cost. Incorporating the congestion cost in the problem introduces non-linearity in the objective function of the interdiction problem, which makes the problem challenging to solve. To address this, we propose two alternate exact solution approaches. The first approach is an inner-approximation-based approach (IBA), which overestimates the convex non-linear objective function and provides an upper bound. A lower bound is obtained from solving the lower-level problem exactly corresponding to the upper bound solution. The upper bound is tightened using improved approximation with new points generated in successive iterations. In the second approach (referred to as SBA), the problem is reformulated as a second-order conic program, which can be solved using an off-the-shelf solver. From our computational experiments on benchmark datasets (CAB and AP), we demonstrate the efficacy of both the proposed methods. However, IBA consistently outperforms SBA by a significant margin.

Go to link
Read More

Journal Articles | 2024

Are green and healthy building labels counterproductive in emerging markets? An examination of office rental contracts in India

Anirban Banerjee, Prashant Das, Franz Fuerst

Financial prudence compels businesses to improve their Environmental, Social, and Governance (ESG) performance when the marginal benefits, pecuniary or non-pecuniary, exceed the marginal costs. For many firms, renting green offices is a feasible ESG activity which may increase their willingness to pay higher rents. Analyzing over 17,000 green rental contracts in India between 2010 and 2022, we find that rents in green-labeled assets and those with health certification command significant premiums between 4 and 21%. However, green rents increased much faster compared to their non-green counterparts, and the propensity to rent green varies significantly across industry segments. We further examine how the market for green offices evolved after a mandatory ESG Disclosure Requirement was enacted in India in 2021. We find that suppliers (landlords) benefited from the regulation by disproportionately increasing rental rates. Existing tenants and foreign firms ended up paying higher rental prices while most other firms, including the assumed target groups of the new policy, redirected their green commitment away from green buildings. Although the policy may yield more positive results in the longer run, a reduced propensity to rent green offices is the opposite of what the ESG Disclosure Requirement tried to achieve.

Go to link
Read More
IIMA