Enhancing Marketing Personalized Shopping Recommendations in the UAE: Leveraging Logic Mining and Advanced Technologies

Personalized shopping recommendations involve offering customized product suggestions to individual users, considering their preferences, behaviors, and past interactions with the platform (Fayyaz, et al., 2020; Sharma, Rana and Kumar, 2021). Amidst the current digital environment, characterized by abundant choices, tailored recommendations are crucial in helping users find suitable products and enhancing their overall purchasing process as fast as possible (Ko, et al., 2022). Customized recommendations are important because they strengthen user involvement and satisfaction, increasing sales and revenue for e-commerce platforms (Tran, et al., 2021).

The significance of personalized purchasing recommendations in e-commerce has grown and become more crucial with the progression of technology (Karthik and Ganapathy, 2021; Alam, et al., 2021). Initially, recommendation systems relied on basic rulebased methods (Aburayya, et al., 2023; Alkashami, et al., 2023). However, machine learning, deep learning, and fuzzy logic have evolved into more advanced and proficient systems capable of delivering exact and pertinent suggestions (Sohn and Kim, 2020; Serrano-Guerrero, et al., 2021). PRs have become a strategic tool for enterprises in the competitive e-commerce industry to distinguish themselves, retain clients, and establish long-term connections (Hallikainen, et al., 2022).

The significance of technology in enhancing personalized shopping suggestions cannot be underestimated (Shwedeh, et al., 2022; Shwedeh, et al., 2023), as it enables the analysis of vast quantities of data, intricate algorithms, and innovative approaches to recommendation modeling (Almahmood and Tekerek, 2022; Zimmermann, et al., 2023). Data mining, semantic web, augmented reality, and artificial intelligence (AI) technologies monitor user behavior, extract valuable insights, and deliver personalized suggestions in real time (El-Deen, et al., 2018; Sharma, Rana and Kumar, 2021). In addition, the integration of explainable AI and user interaction mechanisms enhances the transparency and efficacy of recommendation systems, hence promoting user trust and confidence (Xu, et al., 2020; Ko, et al., 2022). Technology remains crucial in shaping the future of personalized shopping recommendations, fostering creativity, and providing benefits for businesses and customers (Alimour, et al., 2024; Shwedeh, 2024).

Many recommendation systems and methodologies are currently used in personalized purchasing recommendations. These systems strive to optimize the user experience and boost revenue by offering personalized product recommendations. A survey of recommendation systems demonstrates the utilization of diverse models and algorithms for examining user preferences and behaviors (Bleize and Antheunis, 2019; Ko, et al., 2022). These systems utilize machine learning, deep learning, and fuzzy logic to create personalized suggestions by analyzing past user interactions and item characteristics (Fayyaz, et al., 2020; Sharma, Rana and Kumar, 2021). Furthermore, recommendation systems have various applications in various sectors, such as ecommerce, social commerce, health care, and entertainment, underscoring their adaptability and importance in contemporary digital settings (Tran, et al., 2021; Ko, et al., 2022).

Logic mining is becoming increasingly crucial in tailored suggestions since it provides valuable insights into user preferences and behaviors (Sohn and Kim, 2020; Cui, 2021). Logic mining algorithms utilize logical reasoning and pattern recognition to extract significant correlations and rules from extensive datasets. This process aids in identifying pertinent product recommendations (Serrano-Guerrero, et al., 2021; Karthik and Ganapathy, 2021). These algorithms utilize user interactions, such as browsing history, purchase trends, and feedback, to deduce implicit preferences and anticipate future behavior (Sohn and Kim, 2020; Sharma, Rana and Kumar, 2021). In addition, logic mining techniques are being more frequently combined with other recommendation approaches, such as collaborative filtering and contentbased filtering, to improve the accuracy and effectiveness of recommendations (Alam, et al., 2021; Hallikainen, et al., 2022).

Integrating logic mining into PR systems yields several advantages such as higher recommendation precision, improved user contentment, and increased sales revenue (Yang and Yan, 2011; Xu, Ferwerda and Lee, 2020). Using logical thinking and insights derived from data, these systems may effectively address obstacles such as limited data availability, uncertainty in user preferences, and cultural differences (Shwedeh, et al., 2020; Ravikumar et, al., 2023). As a result, they can provide users with more pertinent recommendations tailored to their individual needs and preferences (Karthik and Ganapathy, 2021; Alam, et al., 2021). In addition, combining logic mining with explainable AI allows users to comprehend the reasoning behind recommendations, promoting confidence and transparency in the recommendation process (Almahmood and Tekerek, 2022; Zimmermann, et al., 2023). Logic mining offers a promising solution for dealing with the difficulties of tailored buying recommendations and improving the user experience in e-commerce (Shwedeh, et al., 2020; Ravikumar, et al., 2023).

The intricacies of the e-commerce landscape give rise to numerous challenges in tailored buying recommendations. Data sparsity and cold-start difficulties are significant obstacles in this context (Almahmood and Tekerek, 2022; Zimmermann, et al., 2023). The efficacy of recommendation systems is impeded by the scarcity of user data and the challenge of creating tailored recommendations for new users or items. In addition, the suggestion process is further complicated by the ambiguity of user preferences (Xu, Ferwerda and Lee, 2020; Almahmood and Tekerek, 2022). Comprehending and forecasting user preferences in the face of diverse individual likes and preferences poses a daunting problem for recommendation systems.

In addition, the presence of cultural diversity and market dynamics introduces an additional level of intricacy to tailored buying recommendations (Hallikainen, et al., 2022; Zimmermann, et al., 2023). In regions characterized by many cultural origins, such as the UAE, it becomes progressively challenging to accommodate the distinct preferences and behaviors of various consumer segments. The issues are further exacerbated by the rapidly changing consumer behaviors, which require regular adjustment of recommendation algorithms due to shifting trends and preferences (Sohn and Kim, 2020; Alam, et al., 2021). To stay current and successfully provide individualized purchasing experiences, recommendation systems must be adaptable and responsive to evolving customer habits.

To tackle these ongoing issues, it is necessary to employ inventive strategies and robust frameworks that consider the ever-changing character of e-commerce settings (Almahmood and Tekerek, 2022; Chakraborty and Paul, 2023). By using sophisticated technologies such as explainable AI and augmented reality, the precision and pertinence of PRs can be improved (Zimmermann, et al., 2023). Moreover, it is imperative to cultivate cooperation among industry players and researchers to create customized solutions that address personalized shopping recommendations’ distinct requirements and difficulties (Dahu, et al., 2022; Khadragy, et al., 2022; Shwedeh, Aburayya, et al., 2022). E-commerce platforms can seize new chances for growth and innovation by providing individualized shopping experiences and directly addressing and resolving these difficulties (Shwedeh, et al., 2022; Shwedeh, et al., 2023).

Various variables impact customers’ inclination to purchase tailored shopping recommendations. Product attributes, such as price, brand, and quality, substantially impact customers’ buying choices (Chang and Wildt, 1994; Lee, Cheng and Shih, 2017; Sohn and Kim, 2020). The cost of a product directly influences consumers’ intention to acquire it since they typically evaluate the product’s perceived value concerning its price (Chang and Wildt, 1994). Furthermore, the reputation of a brand and the perceived quality of its products significantly impact how customers perceive the value and trustworthiness of the brand, which in turn affects their intentions to make a purchase (Lee, Cheng and Shih, 2017). Furthermore, user engagement metrics, such as click-through rate and dwell duration, are markers of customer interest and involvement, impacting purchase intention (Shareef, et al., 2008; Ma, et al., 2020).

Trust and satisfaction are crucial in influencing purchase intention in personalized shopping recommendations (Lin and Lu, 2010; Chiu, et al., 2012). Consumers’ confidence in the transaction process is enhanced by their trust in the e-sellers’ reputation and the overall business image, increasing their purchase intention (Shareef, et al., 2008). In addition, consumer trust and the probability of future purchases are increased by both contentment with past purchases and positive word-of-mouth recommendations (Lin and Lu, 2010). Furthermore, the impact of trust on the intention to make repeat purchases online is regulated by the habits developed as a result of repeated pleasant experiences. This highlights the significance of trust in motivating the intention to make purchases (Chiu, et al., 2012).

Furthermore, personalized buying recommendations are influenced by social media interactions and elements related to the source and substance of the recommendations (Onofrei, Filieri, and Kennedy, 2022). The level of consumer engagement and intentions to make online purchases are influenced by the relationships they develop and the emotional connections they establish through social commerce platforms (Ma, et al., 2020; Qi, Yu, and Ploeger, 2020). In addition, the intention to acquire health-care apps is influenced by consumption values, representing consumers’ judgments of the usefulness and advantages of these applications (Chakraborty and Paul, 2023). Thorough comprehension of the factors that influence consumers’ intention to purchase emphasizes the intricate nature of consumer behavior in personalized shopping suggestions. It reinforces the significance of addressing various variables to enhance the effectiveness of recommendation systems.

The following research framework that led to the research hypotheses is given based on the relevant articles and publications reviewed for the issues under investigation (Figure 1).

Figure 1.

Research framework

(Source: Authors’ own research)

Social Exchange Theory offers a conceptual framework for comprehending the dynamics of interpersonal relationships and interactions in many situations, such as online platforms and e-commerce environments. Liao, et al. (2021) and Akarsu, et al. (2020) suggest that people participate in social interactions online to maximize benefits and minimize drawbacks. These factors play a crucial role in the decision-making process in online environments. This idea suggests that individuals are driven to participate in mutual behaviors with the anticipation of gaining advantages or rewards from their contacts (Davlembayeva, et al., 2020; Ferm and Thaichon, 2021). Within the realm of e-commerce, individuals engage in social commerce platforms and online communities with the anticipation of acquiring important insights, endorsements, and assistance from fellow users (Zhang and Liu, 2022). Sharing information and resources fosters trust, dedication, and allegiance among users (Almahmood and Tekerek, 2022; Alimamy and Gnoth, 2022).

Comprehending the fundamental principles of Social Exchange Theory is essential for examining the behaviors and preferences of customers in online environments, specifically concerning tailored buying recommendations. Researchers can enhance recommendation systems in social commerce platforms by analyzing the aspects that impact users’ involvement, trust, and satisfaction

This analysis allows for the development of more efficient systems that cater to individual users’ specific needs and preferences (Khanal, et al., 2020; Ko, et al., 2022). Moreover, Social Exchange Theory emphasizes the significance of reciprocity and mutual advantage in online interactions. This can guide the development and execution of recommendation algorithms prioritizing user-centric experiences (Sharma, et al., 2021; Stalidis, et al., 2023). Researchers can improve the accuracy and significance of tailored recommendations in e-commerce by utilizing Social Exchange Theory insights. This can ultimately enhance user happiness, engagement, and purchase behavior (Tran, et al., 2021; Tian, et al., 2023).

A total response of 280 surveys were gathered throughout the period of 5 months. Since the SEM tool was used to analyze the inferential part of the questionnaire, and the SEM tool can analyze non-normal data, we did not assess the presence of outliers thoroughly. However, we checked if there was any outlier because of data imputation, but none was found. Demographic data presents the data characteristics. This helps us understand the hidden factors that the respondents possess. The survey data shows that most respondents were females 66.07% (185), while their male counterparts had 33.93% (95) responses. Moreover, 46.43% (130) were of Arab origin, 16.07% (45) were of Hispanic origin, 28.57% (80) were of Asian descent, and the remaining 8.93% (25) were of African descent. In addition, all respondents 100% claimed they loved shopping online and offline and received promotional emails because they had store club cards.

We adopted a SEM analytical tool to analyze the inferential statistics of the collected data. Using the SEM analysis tool, we assessed two main models: measurement and structural modeling. Under the measurement model, we assessed parameters that included construct, convergent, and discriminant validity by checking if the Average Variance Explained for each construct was greater than 0.5 (Hair Jr, et al., 2021; Rönkkö and Cho, 2022). Likewise, the composite reliability (CR) threshold should be greater than 0.7, and item loadings should be greater than 0.6 (Ismail, et al., 2020; Sovey, Osman and Mohd-Matore, 2022). For discriminant validity, the Heterotrait Monotrait (HTMT) correlation matrix between constructs should have a maximum threshold of less than 0.9 (Henseler, Ringle and Sarstedt, 2015).

Furthermore, to ascertain that the model is free from collinearity issues, Variance Inflated Factors (VIF) was observed. According to Kim (2019), a model is said to be free from collinearity issues if the VIF value is less than 5. The outcome of the measurement model process is presented in Figure 3, Table 1.

Figure 3.

Measurement model

(Source: Authors’ own research)

As found in Table 1 and Figure 3, the conditions for the construct and the convergent validities were met. Meanwhile, a critical look into our research model shows that our model is a bit complex, having higher order constructs. Given this, it should be recalled that we adopt a reflective—reflective model approach; therefore, Table 1 presents only the measurement model for the first-order constructs. Given this, we calculate the higher-order constructs AVE (1) and CR (2) using the formulae given below: 1AVE=∑1Mli=12/MAVEChallenge =0.9332+0.90122=1.68232=0.841AVEPro Att =0.8522+0.9642+0.92023=2.50163=0.834AVETrust and Sat =0.6912+0.8992+0.80223=2.023=0.673AVEPR=0.8652+0.9422+0.90023=2.5623=0.854

and

2CR=∑i=1nl/nCRChallenge =0.933+0.9012=1.8342=0.917CRPro ATT =0.852+0.964+0.9203=2.7363=0.912CRTrust and Sat =0.691+0.899+0.80232.5333=0.844CRPR=0.865+0.942+0.9003=2.7073=0.902

After calculating AVE and CR for the higher-order constructs, we discovered that the minimum threshold stated for the first-order constructs is fulfilled.

Table 2 presents the HTMT correlation matrix. According to Hensler, et al. (2015), the maximum construct correlation threshold should be less than 0.9. Insight into Table 2 shows that this condition is met. Therefore, convergent validity is ascertained.

Furthermore, we assessed the model for the presence of collinearity. Indeed, a model contains collinearity issues if the VIF is greater than five. Insight into Table 3 shows that the model is free from collinearity issues; thus, we proceed to statistically observe the relationships that exist between the variables under investigation.

Before testing the research hypotheses in this investigation, we assessed the model variance explained using r². The r² reveals a value of 0.344. This implies that the investigated constructs explained a 34.4% variance in PRs. In addition, each exogenous variable’s contribution was examined on the endogenous variable in the form of effect size (f²). The Cohen’s (1988) criterion was adopted in this regard. The SEM analysis presents the effect sizes for challenges, products’ attributes, and trust and satisfaction as 0.358, 0.181, and 0.114, respectively. This suggests, given Cohen's (1988) premise, that the variables under consideration are large, medium, and low, respectively.

As seen in Table 4 and Figure 4, The first hypothesis demonstrates that perceived problems (challenges) faced by shoppers significantly influence PR having (ß = 0.48, T = 5.332), p < 0.05. Hence, we accept the first hypothesis, establishing a significant association between shoppers’ challenges and personal recommendations. Findings suggest that perceived problems in PRs connect with research on consumer behavior and decision-making processes (Bleize and Antheunis, 2019). Consumers generally rely on individualized recommendations to navigate the numerous choices in the marketplace, especially when they experience perceived problems such as information overload or decision complexity (Srikumar and Bhasker, 2005).

Figure 4.

Structural model assessment (hypothesis testing)

(Source: Author’s own research)

Similarly, product qualities (Pro att) strongly influence personalized suggestions, having (β = 0.327, T = 3.449), p < 0.05. Similar to the first hypothesis, we declare in this inquiry that consumers considered a solid, substantial relationship exists between product attributes’ influence on tailored suggestions. The findings corresponded with studies highlighting the role of product attributes in affecting purchase intentions (Chang and Wildt, 1994; Lee, et al., 2017; Sohn and Kim, 2020). Given this, consumers tend to choose tailored recommendations that fit their preferences about product features not limited to price, packaging, quality, and functionality (Lee, et al., 2017; Sohn and Kim, 2020).

Likewise, the findings in this study revealed that perceived trust and satisfaction strongly influence personalized suggestion having (β = 0.102, T = 2.115), p < 0.05. The findings of this study connote the role of trust building and customer satisfaction in e-commerce situations (Shareef, et al., 2008). Trust in recommendation systems is vital for consumers to trust the relevance and reliability of the ideas supplied (Chiu, et al., 2012). In addition, satisfied customers are more likely to connect with and act upon individualized recommendations, resulting in greater purchase intention (Sohn and Kim, 2020).