Intimate partner violence (IPV) is a distinct and concerning facet of domestic violence that has garnered global attention due to its alarming prevalence and severe consequences. The term, "Intimate Partner Violence," encompasses a range of harmful actions within intimate relationships, including physical, sexual, and psychological harm. To be precise the definition include physical assault, sexual coercion, psychological abuse, and controlling behaviour, extending its definition to both current and former partners(1). IPV can be further categorized into three levels: Level I abuse (pushing, shoving, grabbing, throwing objects to intimidate, or causing damage to property and pets); Level II abuse (kicking, biting, and slapping); and Level III abuse (use of a weapon, choking, or attempting strangulate).(2)
The worldwide prevalence of IPV demands immediate attention and concerted efforts to address this pervasive issue. According to a report by the World Bank released on International Women's Day, an astonishing 736 million women aged 15 and older have experienced IPV or nonsexual violence at least once during their lifetime. Shockingly, more than one in every four women has endured physical or psychological abuse at the hands of their partners. The report also reveals that 60% of married women have faced domestic violence in general. (3)
While these global statistics are deeply concerning, the situation in India, a developing nation grappling with challenges such as education, gender inequality, unemployment, and deep-rooted patriarchal norms, is even more distressing. Although there has been a slight improvement in estimates from the NFHS 4 to the NFHS 5 reports, there is still a long journey ahead to fully mitigate this disturbing issue. The percentage of women experiencing spousal violence decreased from 31.2–29.3%. However, the prevalence of physical violence during pregnancy remains 3.1% among women aged 18 to 29 years, with 1.5% of women encountering sexual violence before turning 18.
Recently, the COVID-19 pandemic has exacerbated domestic violence and IPV. Data from the National Commission for Women (NCW) revealed a doubling of domestic violence complaints following the nationwide lockdown in India in 2020. Various police stations reported a sharp increase in calls related to domestic abuse during this period. (4)
While being a victim of IPV has a significant impact on the mental and physical health of the individual, the effect is not limited to the victim herself. IPV is one of the imperative cause of maternal and child mortality. The magnitude of abuse women endure extends beyond their own lives to affect their children, both born and unborn. The physical consequences range from unwanted pregnancies, abortions, and miscarriages to coronary heart disease and chronic back pain. The risk of developing sexually transmitted disease (STD) also increases tenfold for victims who have been sexually abused(5). Exposure to IPV is closely linked to underimmunization of children(6); a higher prevalence of stunting, wasting, and underweight children(7); and an increased risk of diseases such as diarrhea, fever, and acute respiratory infections in children under five years of age, leading to a higher infant mortality rate(8).
To effectively combat IPV, it is crucial to understand its underlying determinants and root causes. The instances of IPV are not discrete or instantaneous but rather are the results of prolonged and deeply ingrained behavioral and observational habits. In most developing countries, gender-based power imbalances are the cornerstone of IPV(9). Households with a history of past domestic abuse are more likely to experience intimate partner violence, highlighting the cyclical nature of this issue. Certain religious and social beliefscondone ideas of limiting women’s participation within the four walls of households, and discarding women’s working rights are likely to fuel the tendency toward IPV(10). The risk of IPV is also greater where violence is generally accepted as a practice for resolving any sort of conflict.(11)
IPV is influenced by multiple factors at different levels, including individual characteristics, personal relationships, community dynamics, and societal norms(12) It can occur across diverse social settings, regardless of factors like religion, caste, or financial status. Previous research has identified several determinants, such as higher educational attainment among women, greater parity, and women engaged in manual labor, as significant characteristics of women associated with the occurrence of IPV. Conversely, according to previous studies, husbands’ demographic traits, including age; social attributes, such as education level; and habits, such as smoking and drinking, influence the occurrence of IPV (13–21).
However, comprehensive research that combines indicators from all four of these domains is lacking. Moreover, the majority of previous research on IPV and its various facets has relied on conventional regression methods. While these methods have strengths, they also have limitations that newer approaches such as machine learning can overcome. For example, traditional methods are unable to address the issue of unspecified interrelationships among the various factors examined in the study.
Moreover, the traditional methods employed in public health research often involve standard hypothesis-driven epidemiologic and demographic analyses. These conventional approaches are designed to focus on a specific set of variables and hypotheses to reduce the number of potential correlates tested. This reduction is performed to mitigate the problem of multiple comparisons, which, if not controlled for, can lead to the identification of spurious significant findings.
Machine learning techniques can overcome these constraints by employing a range of statistical, probabilistic, and optimization methods to uncover concealed and intricate patterns and associations within the data. These techniques empower ML models to delve deep into datasets, uncovering intricate, concealed patterns and nuanced relationships that may elude traditional analytical approaches. In recent years, there has been a surge in the adoption of ML for creating algorithms known for their superior predictive accuracy when compared with conventional methods. This trend highlights the growing recognition of ML as an alternative method for redefining the area of data-driven decision-making and problem solving across various domains(22–29)
However, studies on the adaptation of ML, particularly in the research area of IPV, are scarce. Hence, to address this research gap, this paper will use traditional methods such as descriptive statistics and logistic regression along with algorithm-based models to predict the prevalence of IPV.
Numerous studies have indicated that a woman's age can affect her likelihood of experiencing domestic violence. In general, it is hypothesized that the risk of experiencing violence increases with age because the longer a woman has been married or in a relationship, the more she has been exposed to the risk of domestic violence. However, the relationship between a woman's age and her experience of IPV is not straightforward, as it does not increase linearly and instead fluctuates unpredictably within a relatively narrow age range (30, 31). Being married at a young age is commonly associated with a greater risk of experiencing domestic violence. This belief can be explained from both societal and individual viewpoints. On a societal level, the age at which women marry often reflects their social status. In societies where women have a lower status, there is often a greater incidence of early marriages, which can be linked to an increased vulnerability to domestic violence. On an individual level, a woman's age at marriage can influence her proneness to violence. When a woman marries at a young age, she may not have had the opportunity to develop the understanding and maturity necessary to ensure her security within the marriage. (31, 32)
Several studies have also suggested a connection between vulnerability to domestic violence and the total number of children born to married women. One possible explanation for this association is that as the number of children increases within a family, economic insecurity and limited resources may arise, potentially leading to heightened stress for the head of the household. This elevated stress could contribute to incidents of violence. Therefore, it is posited that a greater number of children might increase the risk of domestic violence (33, 34). Education has played a pivotal role in empowering women by equipping them with the knowledge and skills to access information, navigate the complexities of the modern world, and protect themselves from various forms of violence. It is suggested that women with higher education levels possess greater abilities to safeguard themselves, particularly in situations involving abusive partners. Consequently, women with more extensive education are likely to experience lower levels of violence (31). Religious beliefs can influence individuals’ chances of experiencing domestic violence. However, it is challenging to formulate precise hypotheses regarding the specific impact of religion on the likelihood of domestic violence (35). Furthermore, in some studies, it has been shown that ethnicity can be linked to her likelihood of encountering domestic violence. In India's socioeconomic structure, the most marginalized and disadvantaged groups include scheduled castes, scheduled tribes, and other backward classes, and these groups typically face challenges such as larger family sizes, lower per capita income, and inadequate resources, which can increase the incidence of IPV (36). Women's economic characteristics play a significant role in their empowerment and vulnerability to domestic violence. Economically independent women, often through paid employment, tend to have more influence over financial and household decisions. Consequently, women who are employed are generally at a lower risk of experiencing domestic violence (37). However, it is important to note that during the transition toward economic autonomy, when control over finances shifts from men to women, there can be a period of adjustment that may lead to an increase in incidents of violence against women.
To gain a comprehensive understanding of domestic violence, it is essential to analyze the background characteristics of husbands. According to a multicountry study examining the prevalence and incidence of domestic violence, there is a hypothesis that suggests a negative and consistent relationship between a husband's level of education and the occurrence of violence. However, it is important to note that this association may not be true in all cases. For instance, there appears to be a positive and consistent relationship between a husband's education and the likelihood of violence (31). Research indicates that in relationships where men possess higher levels of education than women, both in terms of their inherent (related to gender) and acquired (linked to educational attainment) status, there is a greater likelihood of them exerting unequal and potentially violent power dynamics within the relationship. Additionally, various studies have proposed that when women surpass their husbands in terms of educational achievement, this can lead to increased susceptibility to marital conflicts (31, 38). Numerous studies have established a clear link between husbands' alcohol use and heightened family-level stress, as well as intimate partner violence (IPV), in women in India. Notably, approximately one-third of Indian women have encountered some form of IPV during their lifetimes. In various Indian contexts, husbands' alcohol consumption has consistently emerged as a significant risk factor for IPV. Moreover, the impact of alcohol use by Indian men extends beyond just IPV, negatively affecting overall family well-being by contributing to health issues such as mental health disorders and injuries. It also has broader social and economic repercussions, impacting the welfare of both immediate and extended family members (7).
Household characteristics, including the household's location (urban or rural) and standard of living index, typically have a significant impact on a woman's life. In the literature on domestic violence, there is a common assumption that women from lower socioeconomic backgrounds are more likely to experience violence than women from higher socioeconomic strata are. However, it is important to note that the low economic status of a household may not directly cause domestic violence but is generally believed to increase risk. Additionally, this relationship is not unidirectional, as the perpetuation and experience of domestic violence can also contribute to worsening economic instability. Women with higher education and from higher income groups may be less likely to disclose their experiences of domestic violence (7, 36).
In Indian society, a prevalent framework is characterized by patriarchal structures and patrilineal kinship systems. Within this context, women are often perceived as "inferior," and husbands may assume an inherent right to dominate and control their wives. These imbalanced power dynamics impede women's involvement in household decision-making, curtail their freedom of movement, and restrict their access to economic resources. Consequently, these circumstances expose women to an elevated risk of facing domestic violence (31). Factors that increase a woman's vulnerability to spousal violence are not solely based on the wife's and her husband's background characteristics. It is also influenced by how well these socioeconomic and cultural traits align with each other. Applying the concept of status inconsistency theory to marital issues suggests that when two individuals with incompatible ascribed or achieved statuses enter into a marital relationship, it can create tension and potentially lead to dissatisfaction within the marriage (36). Generally, significant differences in education level or age between spouses, particularly when the husband is considerably older than the wife or the husband is more educated than the wife, signify power imbalances within the relationship. This is often related to cultural norms that prioritize seniority and masculinity, placing wives who are younger than their husbands in a comparatively disadvantaged position (7).
In general, there is a belief that urban living women are strongly associated with a high risk of domestic violence due to less social interaction. In a multicountry study focusing on the developing world, six out of nine countries (Cambodia, Colombia, the Dominican Republic, Nicaragua, Peru, and Zambia) found that women in urban areas were significantly more likely to report domestic violence than were their rural counterparts (31). Furthermore, several studies have also addressed regional variations in the experience of physical and sexual intimate partner violence (IPV) victimization among married Indian women (39).
The objectives of this paper are twofold. First, we aimed to create models that can predict whether women are at risk of experiencing IPV efficiently. Second, we aimed to determine what important factors are linked to the risk of IPV among these women. To do this, we use data from a large group of people to create a model that can help us predict who might experience IPV. We use different types of machine learning models, such as decision trees, random forests, gradient boosting, and the usual logistic regression model, to make these predictions. The models we developed in this study can be useful for pinpointing groups of women who are more vulnerable to IPV, making it easier to provide them with the right help and support.