Abstract
Introduction
Attempts have been made in previous studies to establish various relationships between some demographic variables of drug users and traffickers and categories of drug offences. Demographic attributes such as gender and age have been studied as they relate differently to drug use and drug trade. For instance, a survey conducted in 2005 by the European Monitoring Centre for Drugs and Drug Addiction (2005) found that the number of male drug users and male clients attending drug rehabilitation services outnumber their female counterparts. While the prevalence of male drug users surpasses that of the females among adults, the lifetime prevalence for ecstasy drugs and cannabis among female students was almost the same with that of the male students (European Monitoring Centre for Drugs and Drug Addiction, 2005). In another study, the 1999 National Household Survey on Drug Abuse consisting of more than 25,000 participants found that males abuse drugs more than females (National Institute on Drug Abuse ([NIDA], 2000). The difference in the rates of drug abuse among males and females was attributed to the opportunity to use drugs. According to NIDA (2000), males usually have more opportunities to use drugs and provided these opportunities were equal, both genders would abuse drugs at the same rate. The findings were said to be applicable to cocaine, marijuana, heroin, and hallucinogens (NIDA, 2000).
Furthermore, the level of involvement in drug trafficking by both males and females has been investigated in previous studies. A study by Li and Feigelman (1994) involving a sample of 351 African American youths, living in low-income settlements, examined the involvement of both genders in drug trafficking. The study found that boys engaged in drug trafficking for the purpose of settling their personal economic needs while for girls, their personal feelings had a strong association with the intention to involve in the trade. Further, the perceptions that other family members, friends, or neighbors were involved in drug trafficking had strong correlation with previous or intended involvement in drug trafficking among both boys and girls (Li & Feigelman, 1994). A related study by Rodriguez and Griffin (2005) examined drug market activities and how such activities differ by gender. The research data consisted of a sample of 129,189 adult males and 24,575 adult females arrested for various drug offences between 2000 and 2003 in the United States. The findings showed that male and female suspects use different tactics to obtain drugs. For instance, females are more likely to obtain drugs in exchange for sex than males, while males are more likely than females to obtain drugs on credit for onward sale. Further, the research held that the effect of gender on how drugs are acquired from sales points is not constant for all drugs including marijuana, methamphetamines, crack, heroin, and cocaine (Rodriguez & Griffin, 2005). A report by UN Women (2014) indicated that women are increasingly getting involved in the drug trade all over the world. This is more common among the low-educated women, those with little or no economic opportunities or those who have suffered (or are suffering) abuse. According to the report, majority of women involved in the drug trade in Latin America are those who were previously involved in either domestic labor or prostitution or both (UN Women, 2014). Anderson and Kavanaugh (2017) investigated the involvement of women in the United States’ new drug market and how such involvement differs from that of men. The study held that women, men, and gender interact in different ways to shape the new drug market, and these interactions are guided by socioeconomic, ethnic, or social status. According to the study, it is erroneous to assume that drug trafficking is a male-dominated endeavor and that women’s participation is limited to objects of exploitation by men (Anderson & Kavanaugh, 2017).
A study by Ludwick, Murphy, and Sales (2015) involving a sample of San Francisco Bay Area women who engaged in illicit drug sales between 2006 and 2009 identified three categories among the participants. The first category consisted of women who fronted their femininity as a shield from police harassment with the notion that women are usually not suspected of drug trafficking. The second category displayed masculine abilities to shield themselves from male harassments and threats. The third category consisted of women who switched between feminism and a display of masculine abilities as the situation demands (Ludwick et al., 2015). Using the two-faceted tactic, this group of women could sell different kinds of drugs and penetrate very challenging environments.
From the foregoing, it is evident that both genders are involved in the use and trade of illicit drugs. It is also clear that the adolescents, youths, and the old are involved with different kinds of illicit drugs either as users or sellers. Even though this knowledge is available in the public domain, attempts have not been made to formalize the same in the form of association rules. The authors identified the need to further study the interplay among gender, age, drug types, and drug offences to arrive at specific rules for the relationships. Therefore, the objective of this study was to employ association rules mining to extract specific rules that establish how gender, age groups, drug types, and categories of drug offences interrelate. The established rules should serve as useful tools for guide and reference purposes in the global fight against drug use and trafficking.
Application of Data Mining and Machine Learning Approaches in Drugs Research
Data mining and machine learning techniques have been employed in the research on drug use and trafficking. One of such studies is authored by Jimenez, Anupol, Gajal, and Gervilla (2018) where various machine learning algorithms were deployed to study the motives behind substance use amongst students. The research data were drawn from a sample of 9300 students aged between 14 and 18 years, across 22 schools on the Mallorca Island. Five learning algorithms, namely decision tree, artificial neural network, logistic regression, naïve Bayes, and k-nearest neighbor were deployed. The data consisted of four drug/substance types, namely alcohol, tobacco, cocaine, and cannabis. The result of the analyses indicated that the most frequent motives for substance use are for pleasure, because friends consume, to feel new sensations, and to forget life challenges (Jimenez et al., 2018).
In another study, Zhou, Sani, and Luo (2016) employed machine learning techniques to analyze text and image data from Instagram to discover patterns of drug use and the demographics of drug users. Drug-related posts on Instagram were obtained by identifying the hashtags accompanying the posts, and image analysis was conducted to identify the gender and probable ages of the posters. Furthermore, location and time analysis were conducted to identify the location and the time of the day during which the drug-related post was made. The results showed that the most abused drugs are cannabis, cough syrup, and prescription pills such as Vicodin. Further, the results identified two peak hours of cannabis consumption among users as 16:00 and 21:00 hr, irrespective of time zone. Additionally, the results revealed that majority of drug users who posted on Instagram are aged between 20 and 40 years; and most of them have common interests in certain comedians, music, and celebrities (Zhou et al., 2016).
A related study by Chou, Chang, and Puspitasari (2020) deployed the text mining technique to review the trend in drug abuse research. The study analyzed a total of 19,843 drug-related papers from PubMed using text mining. The study found that the most widely used questionnaires in drug research include the Addiction Severity Index (ASI) and the Addiction Research Center Inventory (ARCI), developed in 1980 and 1966 respectively. Furthermore, the study found that current research on drugs do not only focus on evaluating the extent of drug addiction but also pays attention to other parameters such as quality of life, the physical, mental, psychiatric, and sleep status of the drug addicts (Chou et al., 2020).
In their study, Li, Xu, Shah, and Mackey (2019) developed and evaluated a deep learning approach that can detect when Instagram is used for drug-related transactions. The authors used Web scrapper to retrieve 12,857 drug-related posts between July and October 2018. Then, the patterns inherent in the drug transaction-related posts were examined using the long short-term memory unit in the recurrent neural network. The performance of the novel method was compared against three existing models, namely random forest, support vector machines, and decision tree. According to the authors, the novel method performed well in the detection of online illicit drug transactions (Li et al., 2019).
Atsa’am et al. (2022) developed a predictive model for classifying drug suspects and offenders as either drug peddler or non-drug peddler. The authors used a secondary dataset comprising of records on 262 arrestees of various drug offences and misdemeanors in central Nigeria to construct the model using artificial neural network. The model requires parameters such as suspect’s age, type of substance involved, weight of the substance, and gender of the suspect to make a prediction whether the suspect is a drug peddler or non-drug peddler. According to Atsa’am et al. (2022), the model has a classification accuracy of 83% and is appropriate for use at ports of entry and exit of a country to classify apprehended drug suspects.
The reviewed literature show that some aspects of data mining and machine learning have previously been applied in the research on illicit drugs. However, there seems to be no study that applied association rules mining to research on the drug attributes that frequently co-occur. Therefore, the goal of this study was to conduct association rules mining on a secondary drug data. The research extracted rules that show the interrelationships among illicit drug attributes. The rules subgroup the global drug suspect base into segments that can guide how drug intervention and control programs are formulated and implemented.
Materials and Methods
Data
Data Variables (Atsa’am et al., 2022).
Demographic Breakdown of Arrestees.
Concept of Association Rules
Association rules mining is a machine learning technique used to extract hidden associations or co-occurrences among dataset variables (Mahmood, Shahbaz, & Guergachi, 2014). The technique is built on the theory that if an antecedent 
Data Preprocessing
The variable, Exhibit weight, is not relevant to the research focus and was eliminated during data preprocessing. Association rules mining requires data in transaction format prior to rules extraction. The data were converted to transaction format by defining separate variables for each category of Exhibit type, Gender, and Group. That is, the new data variables were psychotropic substance, cannabis sativa, male, female, drug peddler, and non-drug peddler. Furthermore, the Age variable was split into two new variables as “Youth” and “Older adult”. Youth included arrestees with ages ranging from 14 to 24 years while older adults were defined from 25 to 63 years.
For each record, the data points were represented by the variable name to indicate the occurrence of that variable in the record. Blank spaces were used to indicate where a variable is not applicable to a record. This effectively converted the research data to a transaction format shown in the Appendix.
Association Rules Extraction
Association Rules on Drug Variables.
Results
Table 2 describes the demography of the arrestees in terms of gender and age. Furthermore, Table 3 gives the strongest rules that show the co-occurrences among various drug variables. The association rules are generalizable since each produced a high confidence score - between 92 and 100%.
The semantics of the 6 rules are explained in the following.
Rule 1: Possession of psychotropic substances for non-commercial purpose (non-drug peddling) is strongly associated with persons of youthful age who are often males.
Rule 2: Possession of psychotropic substances for commercial purpose (drug peddling) is strongly associated with older adults who are often males.
Rule 3: Possession of psychotropic substances for non-commercial purpose (non-drug peddling) is strongly associated with older adults who are often males.
Rule 4: Possession of cannabis sativa for non-commercial purpose (non-drug peddling) is strongly associated with persons of youthful age who are often females.
Rule 5: Possession of cannabis sativa for commercial purpose (drug peddling) is strongly associated with older adults who are often females.
Rule 6: Possession of cannabis sativa for commercial purpose (drug peddling) is strongly associated with older adults who are often males.
Discussion
Let it be recalled that the variable “Youth” is defined in this study as drug suspects with ages ranging between 14 and 24. In Nigeria where the research data come from, persons within this age brackets are mostly students in secondary or tertiary schools. At this level, majority of these individuals are dependent on their parents or guardians for living expenses and personal needs. Consequently, their involvement with drugs is more likely for personal consumption than for commercial purposes. These explanations are consistent with Rules 1 and 4. Specifically, male youths are strongly associated with the consumption of psychotropic substances (Rule 1) while female youths often consume cannabis sativa (Rule 4). This agrees with the findings of a study by Al-Gburi, Al-Murshedi, Alridha, and Baiee (2020) which found drug use prevalent among secondary school students. Most of the respondents in that study blamed drug use to peer influence and the zeal to experience new sensations.
Rules 2, 3, 5, and 6 are related to older adults defined in this study as arrestees aged between 25 and 63. During these ages, most individuals are financially independent and often under pressure to raise money and cater for their families. To live up to their responsibilities, some are compelled to tow the illegal path of drug peddling. This explains why Rule 2 establishes a frequent co-occurrence among drug peddling, older adult, psychotropic substances, and male. The explanation for Rule 2 is same for Rule 5 with the difference that females are strongly associated with the peddling of cannabis sativa and not psychotropic substances like males. Rule 3 establishes a frequent co-occurrence of psychotropic substances’ use with older adults who are males. This rule is symmetrical with Rule 2 and both rules agree with the findings of Johnson (2003) which held that most sellers of illegal drugs also consume the same and only a few do not consume what they sell.
The findings of the present study have far-reaching implications on how gender issues can be incorporated in the problems of drug abuse, control, and rehabilitation of addicts. A report by UN Women (2014) pointed out that the facilities and personnel in place at some drug treatment or detention centers are not sensitive to gender. Rules 5 and 6 point to the fact that both older-adult males and females have a strong association with the trafficking of cannabis sativa. This implies that the control, intervention, or punitive measures and facilities against the peddling of cannabis sativa should encompass all genders. Furthermore, the extracted rules give an idea on how the drug suspect base can be segmented for appropriate intervention and control programs. For instance, flowing from Rules 1 and 4, the youth of either genders are associated with the consumption of cannabis sativa or psychotropic substances. Consequently, outreach programs against the use of these substances could target secondary schools, university campuses, and night clubs where this segment of drug users can be found in large numbers.
Conclusion
Drug types, drug offence categories, and suspects demographics co-occur in different ways. The analyses in this study considered 2 drug types: cannabis sativa and psychotropic substance; 2 drug offence categories: peddling and consumption; and two suspect’s demographic variables: gender and age. Association rules mining technique was deployed on a dataset of 262 arrestees of various drug offences to examine the patterns of combination among these variables. Six rules were extracted which show the different levels of involvement with drugs among the youth, older adults, males, and females. The rules can be a reference tool for drug suspect base segmentation to ensure that appropriate drug intervention and control strategies are designed and deployed to the appropriate target group. Furthermore, security agents can adopt the rules as a reference tool when dealing with drug suspects and offenders. The study is limited to the variables captured in the research data. For instance, only two drug types, namely cannabis sativa and psychotropic substance were involved. Further, demographic variables such as marital status and race were not examined. In a future study, a dataset consisting of a larger variable set should be considered for generating association rules of how the variables co-occur.
Footnotes
Declaration of Conflicting Interests
Funding
ORCID iD
