Earticle

The proliferation of items increased the difficulty of customers in finding the specific items they want to purchase. To solve this problem, companies adopted recommender systems, such as collaborative filtering systems, to provide personalization services. However, companies use only meaningful and essential data given the explosive growth of data. Some customers are concerned that their private information may be exposed because CF systems necessarily deal with personal information. Based on these concerns, we analyze the effects of the amount of information on recommendation performance. We assume that a customer could choose to provide overall information or partial information. Experimental results indicate that customers who provided overall information generally demonstrated high performance, but differences exist according to the characteristics of products. Our study can provide companies with insights concerning the efficient utilization of data.

This paper presents a novel method for evaluating the usefulness of online product recommendation. Previous studies on evaluating recommendation systems have mostly relied on two methods：testing the accuracy of estimating user preferences by recommendation systems, or empirically testing the effectiveness with lab experiments involving human participants. The former does not measure the usefulness directly and hence can be misleading; the latter is expensive in that it requires a working online store system and test participants. In order to address the problems, the proposed approach uses simulation to imitate customer behavior and evaluate the usefulness of recommendation. Models for user behavior and an abstract Internet store are developed for simulation. Actual simulation experiments are performed to illustrate the use of the approach.

협업 필터링은 사용자의 기록을 바탕으로 선호도를 예측하여 추천하는 방법이다. 최근 협업 필터링 중 머신러닝 알 고리즘을 이용하는 연구가 활발히 이루어지고 있으며 그중에서 오토인코더 모델들의 추천 성능이 뛰어났다. 오토인 코더 모델들은 추천 성능 향상을 위해 별도의 데이터 셋 전처리 과정을 거친다. 하지만 현재 모델들이 사용하는 기 존의 전처리에서는 차원 축소 및 복원으로 인한 정보손실 문제와 데이터 셋들의 불균일한 분포 문제를 해결하지 못 한다. 따라서 본 연구에서는 사용자 기반의 오토인코더 모델에 적합한 전처리 시스템을 제안한다. 본 연구는 필터 모듈과 분배 모듈로 이루어져 있으며 기존의 전처리의 문제들을 순차적으로 개선한다. 필터 모듈에서는 정보손실 문 제를 개선하고 분배 모듈에서는 데이터 셋들의 불균일한 분포 문제를 개선하였다. 그리고 각각의 모듈들을 기존의 전처리 방법과 비교하는 실험을 하여 개선된 결과를 확인하였다. 또한 제안한 방법을 적용한 데이터 셋으로 사용자 기반 오토인코더 모델을 학습시켰을 때 추천 성능이 올라가는 것을 실험을 통해 확인하였다.

Collaborative filtering(CF) is a method of recommending possible preferences of a user based on existing records. Recently, CF research using machine learning algorithms has been actively conducted, among which Autoencoder models have outstanding recommendation performances. Autoencoder models go through preprocessing to improve recommendation performance. However, traditional preprocessing used by current models cannot solve information loss due to dimensional reduction and restoration as well as uneven distribution of datasets. Therefore, in this work, we propose a preprocessing system suitable for user-based Autoencoder models. This work consists of filter modules and distribution modules, which sequentially reduce the problems of traditional preprocessing. We minimized the information loss problems in filter modules and the uneven distribution of datasets in distribution modules. We then conducted experiments comparing each module to conventional preprocessing to confirm the improved results. We also proved that recommendation performance increase when we train a user-based Autoencoder model with datasets that applied the suggested method through an experiment.

Item-based collaborative filtering (IBCF) is an important technology that is widely used in recommender system of online shopping malls. It uses historical information to compute item-item similarity and make predictions. However, in offline shopping each customer’s purchasing pattern can be occurred continuously and repeatedly due to time and space constraints contrast to online shopping. Those facts can make IBCF to have limitations from being applied to offline shopping malls directly. In order to improve the quality of recommendations made by IBCF in offline shopping mall, we propose an ensemble approach that considers both item-item similarity of IBCF and each customer’s purchasing patterns which are modeled by item networks. Our experimental results show that this approach produces recommendation results superior to those of existing works such as pure IBCF or bestseller approaches.

In this paper, the author devised new decision recommendation ordering method of items attributed by age to improve accuracy of recommender system. In conventional recommendation system, recommendation order is decided by high order of preference prediction. However, in this paper, recommendation accuracy is improved by decision recommendation order method that reflect age attribute of target customer and neighborhood in preference prediction. By applying decision recommendation order method to recommender system, recommendation accuracy is improved more than conventional ordering method of recommendation.

In this paper, the author proposed following two methods to improve the accuracy of the recommender system. First, in order to classify the users more accurately, the author used a EMC(Expanded Moving Center) heuristic algorithm which improved clustering accuracy. Second, the author proposed the Neighborhood-oriented preference prediction method that improved the conventional preference prediction methods, so the accuracy of the recommender system is improved. The test result of the recommender system which adapted the above two methods suggested in this paper was improved the accuracy than the conventional recommendation methods.

서비스 업종뿐만 아니라 일반 소비재 산업에서도 빅 데이터 분석 기술을 활용한 개인화 추천 서비스와 같은 CRM 전략의 중요성이 높아지고 있다. 그러나, 상품 추천 모형에 대한 기존 연구는 대부분 추천 적중률을 높이기 위한 머신러닝 모형 개발과 개선에 집중되어 왔기 때문에 추천 활동 이후 고객과의 중장기적인 관계는 고려되지 못했다. 본 연구에서는 이러한 한계를 극복하기 위하여 소셜네트워크분석 (Social Network Analysis) 분석기법을 상품 네트워크 분석 영역에 응용한 PNA(Product Network Analysis) 기법을 협업필터링 모형에 연계하여 단순한 추천 적중률뿐만 아니라 중장기적인 고객의 충성도를 개선할 수 있는 추천 시스템을 제안하고자 한다. CF/PNA(Collaborative Filtering based on Product Network Analysis) 개인화 추천 시스템은 거래 데이터를 기반으로 상품 네트워크를 구성한 후, 다양한 후보 추천상품 중에 PNA 분석 결과를 토대로 중장기적으로 고객 충성도가 향상될 수 있는 상품을 선정하여 최우선으로 추천하는 시스템이다. 국내 대형 아웃도어 기업과의 협력으로 실제 고객 데이터를 기반으로 모델링과 검증을 수행한 결과, 모형의 추천 적중률뿐만 아니라, 거래빈도, 재구매율, 교차구매지수 등 거래의 질적수준을 평가하는 고객로열티 지수도 함께 증가함을 확인하였다. 본 연구는 추천 적중률이 좋아도 상품 간의 연계성이 부족하여 지속적인 고객 관계를 유지하기 어려운 경우 CRM 관점에서는 오히려 추천을 유보하는 것이 바람직할 수 있다는 전략적인 통찰력을 얻을 수 있다.

In the general consumer goods industry as well as the service industry, the importance of CRM strategies such as personalized recommendation is increasing. However, most recommendation studies have focused on enhancing the hit rate of recommendation, not on the long-term customer relationship. To expand the previous relevant studies, we propose a new recommendation system called the CF/PNA (Collaborative Filtering based on Product Network Analysis) which integrates the PNA (Product Network Analysis) technique with collaborative filtering model. The CF/PNA recommendation system configures a product network based on transaction data, and then selects products that can improve customer loyalty in the mid to long term based on PNA analysis results among candidate recommended products. As a result of modeling and verification based on actual customer data in cooperation with outdoor companies, it was confirmed that customer loyalty index such as buying frequency, repurchase rate, and cross-purchase ratio evaluating the qualitative level of purchasing as well as the hit ratio have been increased. This study provides strategic insights that it may be desirable to withhold recommendations from a CRM viewpoint when specific product has not sure to maintain a continuous customer relationship due to lack of linkage between other products

전인적인 인격 형성에 근본적인 역할을 하고 있는 음악과 치료가 융합된 분야인 음악치료는 다양하고 복잡한 치료 방법을 가지고 있다. 음악치료를 담당하고 있는 음악치료사들은 내담자와의 상담에 역전이와 같은 경우의 현상이 발생하기도 하며, 심리적 소진을 경험하고 있기에, 음악 치료의 최종 목표 도달에 많은 어려움이 발생하고 있는 상황이 다. 본 논문에서는 음악치료를 위하여 방문한 내담자와의 원활한 음악 치료 상담을 위하여 협업 필터링 기반의 음악치료 상담 자료 추천 모델을 제안한다. 제안 모델은 기존 상담 데이터와 새로운 상담자의 데이터를 유클리디안 거리 알고리즘 을 통하여 유사도를 파악하고, 이를 통하여 유사 상담 자료를 추천하는 것으로서, 음악치료사들은 음악 치료가 필요한 상담자에게 가장 적합한 상담 자료를 제공할 수 있기에 원활한 상담이 진행될 수 있을 것으로 기대된다.

Music therapy, a field that convergence music and treatment, which play a fundamental role in personality formation, possesses diverse and complex treatment methods. Music therapists in charge of music therapy may experience the same phenomenon as countertransference in consultation with clients. In addition, experiencing psychological burnout, there are many difficulties in reaching the final goal of music therapy. In this paper, we provide a collaborative filtering-based music therapy consultation data recommendation model for smooth music therapy consultation with clients who visited for music therapy. The proposed model grasps the similarity between the conventional consultation data and the new consultant data through the euclidean distance algorithm. This is to recommend similar consultation materials. Since music therapists can provide optimal consultation materials for consultants who need music therapy, smooth consultation is expected.

협업 필터링은 추천 시스템의 대표적인 기술로, 사용자 또는 항목 간의 유사성을 기반으로 추천 대상을 결정하는 방 법이다. 협업 필터링의 계산 속도를 높이기 위해 분산 병렬 협업 필터링이 제안되었다. 그러나 데이터 분산 불균형 으로 인한 데이터 왜곡 문제는 여전히 남아 있다. 따라서 본 연구에서는 데이터 분포 불균형으로 인한 데이터 왜곡 문제를 해결하기 위해 사용자 취향 레퍼토리 분석 방법에 기반한 데이터 분포 및 처리 방법을 제안하였다. 레퍼토 리 분석은 사용자가 이전에 사용했던 서비스 영역의 범위와 수를 분석하는 것이며, 제안된 기법에서는 사용자가 이 전에 사용했던 서비스에 대한 레퍼토리 분석 결과를 기반으로 데이터를 배포한다. 레퍼토리 분석 결과를 기반으로 사용자 데이터를 배포하는 실험이다. 성능은 RMSE 및 실행 시간 측정으로 충분히 사용할 수 있음을 보였다.

Collaborative filtering is a representative technique of recommendation systems. Collaborative filtering is a method of determining a recommendation target based on similarity between users or items. Distributed parallel collaborative filtering was proposed to improve the computational speed of collaborative filtering. However, the data skewness problem caused by imbalanced data distribution still remains. Therefore, in this study, we proposed a data distribution and processing method based on the user taste repertoire analysis method to solve the data skewness problem caused by data distribution imbalance. So, the repertoire analysis is to analyze the range and number of areas of the services previously used by the users. In the proposed method, data is distributed based on the results of the repertoire analysis of the services previously used by the users. Our experiment distributing data of users based on the results of the repertoire analysis. It was shown that the performance was sufficiently usable as measuring RMSE and execution time.

정보 통신과 스마트폰 등의 발달로 인한 편리한 접근성과 다양한 아이템의 종류로 인해 개인 맞춤형 추천의 필요성은 점차 커지고 있다. 날씨 및 기상환경은 사용자의 장소 및 활동의 의사결정에 많은 영향을 미친다. 이러한 날씨 정보를 이용하면 추천에 대한 사용자의 만족도를 높일 수 있다. 본 논문에서는 모바일 플랫폼에서 사용자의 위치 정보에 대한 생활지수를 활용하여 성향이 유사한 사용자를 구하고 장소에 대한 선호도를 예측하여 장소를 추천함으로 써 생활지수를 이용한 협업 필터링 기반 장소 추천 시스템을 제안한다. 제안된 시스템은 사용자의 날씨를 분석하고 분류하기 위한 날씨 모듈과 장소 추천을 위한 협업 필터링을 사용하는 추천 모듈, 그리고 사용자의 선호도 및 후기 관리를 위한 관리 모듈로 구성된다. 실험 결과, 제안된 시스템은 협업 필터링 알고리즘과 생활지수의 융합 및 개인의 성향을 반영하는 측면에서 유효함을 확인할 수 있었다.

The need for personalized recommendation is growing due to convenient access and various types of items due to the development of information communication and smartphones. Weather and weather conditions have a great influence on the decision-making of users' places and activities. This weather information can increase users' satisfaction with recommendations. In this paper, we propose a collaborative filtering-based place recommendation system using living index by utilizing living index of users' location information on mobile platform to find users with similar propensity and to recommend places by predicting preferences for places. The proposed system consists of a weather module for analyzing and classifying users' weather, a recommendation module using collaborative filtering for place recommendations, and a management module for user preferences and post-management. Experiments have shown that the proposed system is valid in terms of the convergence of collaborative filtering algorithms and living indices and reflecting individual propensity.

협력 필터링은 사용자가 선호했던 항목들의 기록을 토대로 항목을 추천하는 방법으로서 상업 사이트에서 매우 널리 사용되어 왔다. 이 방식의 기본 개념은 유사한 사용자들을 찾아서 그들의 평가등급을 통합하여 새로운 항목 추천에 이용하는 것이다. 따라서 유사도의 정확한 측정은 추천 성능에 매우 중요한 일이다. 본 논문에서는 사용자가 과거에 부여했던 평가등급들을 기준으로 하여 상대적으로 각 평가치를 다루는 새로운 유사도 공식을 제안한다. 광범위한 실험을 통해 제안된 공식이 기존 공식들보다 더 신뢰할 수 있음을 밝혔는데, 이는 극단적인 유사도값의 발생이 현저히 감소하였고, 유사도가 큰 이웃들만을 참조하였을 때 성능이 개선되었기 때문이다. 특히 실험 결과, 제안 공식은 평가 범위가 큰 데이터셋에 대해 기존 공식들보다 우수한 성능을 나타냈다.

Collaborative filtering has been most widely used in commercial sites to recommend items based on the history of user preferences for items. The basic idea behind this method is to find similar users whose ratings for items are incorporated to make recommendations for new items. Hence, similarity calculation is most critical in recommendation performance. This paper presents a new similarity measure that takes each rating of a user relatively to his own ratings. Extensive experiments revealed that the proposed measure is more reliable than the classic measures in that it significantly decreases generation of extreme similarity values and its performance improves when consulting neighbors with high similarites only. In particular, the results show that the proposed measure is superior to the classic ones for datasets with large rating scales.

추천시스템은 사용자들의 관심을 끄는 아이템을 그들이 보다 쉽게 찾도록 도와주거나, 그들의 기호에 기반하여 의미 있는 아이템들을 제공한다. 지금까지 가장 성공적이었던 협업 필터링 기반 추천시스템은 다른 사용자들의 의견을 참조하여 추천을 원하는 사용자에게 추천을 한다. 즉, 아이템들에 대한 사용자 기호를 나타내는 다른 사용자들의 평가정보가 추천을 위한 정보원으로 사용된다. 이처럼, 협업 필터링 기반 추천시스템이 사용자들의 기호만을 이용하도록 설계되었지만, 다른 정보를 이용하면 추천시스템의 성능과 정확도를 높일 수 있을 것으로 사료되어, 본 논문에서는 유사 정도와 인구통계학 정보를 이용한 협업 필터링 기반 추천시스템을 제안한다. 이런 추천시스템에서는 평가정보가 계속적으로 누적되기 때문에, 추천시스템의 정확도를 유지할 수 있는 한, 사용하는 데이터의 양을 줄이는 게 중요하다. 본 논문에서는 유사 정도와 인구통계학 정보를, 사용할 데이터의 양을 줄이기 위한 기준으로 사용하여 자연스레 시스템의 성능을 향상시켰다. 본 논문에서는 실험을 통하여 유사 정도의 사용이 추천시스템의 정확도를 높여주었고, 특정 인구통계학 정보의 사용도 추천시스템의 정확도를 높였음을 보였다.

Recommender systems help users find their interesting items more easily or provide users with meaningful items based on their preferences. Collaborative filtering(CF) recommender systems, the most successful recommender system, use opinions of users to recommend for an active user who needs recommendation. That is, ratings which users have voted on items to indicate preference on them are the source for making recommendation. Although CF systems are designed only to use users' preferences as the source of recommendation, use of some available information is believed to increase both the performance and the accuracy of CF systems. In this paper, we propose a CF recommender system which utilizes both degree of match and demographic information(e.g., occupation, gender, age) to increase the performance and the accuracy. Since more and more information is accumulated in CF systems, it is important to reduce the data volume while maintaining the same or the higher level of accuracy. We used both degree of match and demographic information as criteria for reducing the data volume, thereby naturally enhancing the performance. It is shown that using degree of match improves the prediction accuracy too in CF systems and also that using some demographic information also results in better accuracy.

Collaborative filtering is a representative technique of a recommendation system. Collaborative filtering is a method of determining a recommendation target based on similarity between users or items. Distributed parallel collaborative filtering was proposed to speed up the computational speed of collaborative filtering. However, the data skewness problem caused by imbalanced data distribution still remains. Therefore, in this study, we proposed a data distribution and processing method based on the user taste repertoire analysis method to solve the data skewness problem caused by data distribution imbalance. So, the repertorie analysis is to analyze the range and number of areas of the services previously used by the users. In the proposed method, data is distributed based on the results of the repertorie analysis of the services previously used by the users. Our experiment distributing data of users based on the results of the repertoire analysis. It was shown that the performance was sufficiently usable as measuring RMSE and execution time.

The development of computer and information technology has been combined with the information superhighway internet infrastructure, so information widely spreads not only in special fields but also in the daily lives of people. Information ubiquity influences the traditional way of transaction, and leads a new E-commerce which distinguishes from the existing E-commerce. Not only goods as physical but also service as non-physical come into E-commerce. As the scale of E-Commerce is being enlarged as well. It keeps people from finding information they want. Recommender systems are now becoming the main tools for E-Commerce to mitigate the information overload.Recommender systems can be defined as systems for suggesting some Items (goods or service) considering customers' interests or tastes. They are being used by E-commerce web sites to suggest products to their customers who want to find something for them and to provide them with information to help them decide which to purchase. There are several approaches of recommending goods to customer in recommender system but in this study, the main subject is focused on collaborative filtering technique. This study presents a possibility of pre-evaluation for the prediction performance of customer's preference in collaborative filtering before the process of customer's preference prediction. Pre-evaluation for the pre-diction performance of each customer having low performance is classified by using the statistical features of ratings rated by each customer is conducted before the prediction process.In this study, MovieLens 100K dataset is used to analyze the accuracy of classification. The classification criteria are set by using the training sets divided 80% from the 100K dataset. In the process of classification, the customers are divided into two groups, classified group and non classified group. To compare the prediction performance of classified group and non classified group, the prediction process runs the 20% test set through the Neighborhood Based Collaborative Filtering Algorithm and Correspondence Mean Algorithm. The prediction errors from those prediction algorithm are allocated to each customer and compared with each user's error.Research hypothesisTwo research hypotheses are formulated in this study to test the accuracy of the classification criterion as follows.Hypothesis 1: The estimation accuracy of groups classified according to the standard deviation of each user's ratings has significant difference.To test the Hypothesis 1, the standard deviation is calculated for each user in training set which is divided 80% from MovieLens 100K dataset. Four groups are classified according to the quartile of the each user's standard deviations. It is compared to test the estimation errors of each group which results from test set are significantly different.Hypothesis 2: The estimation accuracy of groups that are classified according to the distribution of each user's ratings have significant differences.To test the Hypothesis 2, the distributions of each user's ratings are compared with the distribution of ratings of all customers in training set which is divided 80% from MovieLens 100K dataset. It assumes that the customers whose ratings' distribution are different from that of all customers would have low performance, so six types of different distributions are set to be compared. The test groups are classified into fit group or non-fit group according to the each type of different distribution assumed. The degrees in accordance with each type of distribution and each customer's distributions are tested by the test of 　　 goodness-of-fit and classified two groups for testing the difference of the mean of errors. Also, the degree of goodness-of-fit with the distribution of each user's ratings and the average distribution of the ratings in the training set are closely related to the prediction errors from those prediction algorithms. Through this study, the customers who have lower performance of prediction than the rest in the system are classified by those two criteria, which are set by statistical features of customers ratings in the training set, before the prediction process.

본 논문은 교양과목 추천 시스템 개발을 목적으로, 학생들의 흥미를 기반으로 교양과목을 추천하는 방법을 제안한다. 사용자로부터 흥 미 키워드를 입력받아 관련 키워드를 제시하고, 사용자가 원하는 키워드를 선택한다. 최종적으로 교양과목 추천 알고리즘을 적용하여 점수가 높은 과목을 추천한다. 시스템 개발을 위해, 218개의 인문학 과목을 수집하고 키워드를 추출한 후 워드 임베딩을 하여 키워드간 의 유사도를 비교한다. 이 시스템은 학생들의 적성에 맞춘 과목을 추천하여 교육적 의의를 높이는 데 목적이 있다. 키워드 기반 추천 시스템은 초기 사용자 데이터가 부족한 상황에서도 개인화된 추천을 제공할 수 있으며, 학생들이 단순히 학점을 잘 받을 수 있는 과목 대신, 흥미와 적성에 맞는 과목을 선택하도록 도와준다. 이는 대학 교육의 본래 목표인 전인 교육을 실현하는 데 중요한 역할을 한다.

This study aims to develop a liberal arts course recommendation system based on students' interests. Users input interest keywords, from which related keywords are prioritized for selection. A course recommendation algorithm is applied to suggest courses with high scores. For system development, we collected 218 liberal arts subjects, extracted keywords, and performed word embedding. This system enhances educational significance by recommending courses tailored to students' aptitudes. The keyword-based recommendation system can provide personalized suggestions even with limited initial user data, guiding students to choose courses aligned with their interests and aptitudes rather than merely focusing on grades. This approach plays a crucial role in realizing the holistic education goal of higher education.

Most studies on recommender systems to evaluate recommendation performances focus on offline evaluation methods utilizing past customer transaction records. However, evaluating recommendation performance through real-world stimulation becomes challenging. Moreover, such methods cannot evaluate the duration of the recommendation effect. This study measures the personalized recommendation (stimulus) effect when the product recommendation to customers leads to actual purchases and evaluates the duration of the stimulus personalized recommendation effect leading to purchases. The results revealed a 4.58% improvement in recommendation performance in the online environment compared with that in the offline environment. Furthermore, there is little difference in recommendation performance in offline experiments by period, whereas the recommendation performance declines with time in online experiments.

A single-learner model which integrates the user's positive and negative perceptions is proposed by augmenting counterfactual data to the interaction data between users and items, which are mainly used in collaborative filtering in this study. The proposed CausRec showed superior performance compared to the existing NCF model in terms of F1 value and AUC in experiments using three published datasets: MovieLens 100K, Amazon Gift Card, and Amazon Magazine. Compared to the existing NCF model, the F1 and AUC values of CausRec showed 1.2% and 2.6% performance improvement in MovieLens 100K data, and 2.2% and 10% improvement in Amazon Gift Card data, respectively. In particular, in experiments using Amazon Magazine data, F1 and AUC values were improved by 11.7% and 21.9%, respectively, showing a significant performance improvement effect. The performance of CausRec is improved because both positive and negative perceptions of the item were reflected in the recommendation at the same time. It is judged that the proposed method was able to improve the performance of the collaborative filtering because it can simultaneously alleviate the sparsity and imbalance problems of the interaction data.

In this study, with the goal of developing a deep learning-based product recommendation model for effective matching of influencers and products, a deep learning model with a collaborative filtering model combined with generalized matrix decomposition(GMF), a collaborative filtering model based on multi-layer perceptron (MLP), and neural collaborative filtering and generalized matrix Factorization (NeuMF), a hybrid model combining GMP and MLP was developed and tested. In particular, we utilize one-class problem free boosting (OCF-B) method to solve the one-class problem that occurs when training is performed only on positive cases using implicit feedback in the deep learning-based collaborative filtering recommendation model. In relation to model selection based on overall experimental results, the MLP model showed highest performance with weighted average precision, weighted average recall, and f1 score were 0.85 in the model (n=3,000, term=15). This study is meaningful in practice as it attempted to commercialize a deep learning-based recommendation system where influencer's promotion data is being accumulated, pactical personalized recommendation service is not yet commercially applied yet.

대부분의 추천 시스템 관련 연구는 수집된 데이터를 이용하는 오프라인 평가방식을 통해 추천 시 스템을 구축하고 있으며, 특정 시점에 대해서만 추천 성능을 평가하는 홀드아웃(Hold-Out) 등의 기법을 사용하고 있다. 이와 같은 방식은 실제 온라인 환경에서 고객이 제품을 추천 받은 시점과 제품을 구매하는 시점 간의 시간 차이(Time-Lag)에 따른 추천 효과를 측정하는 것이 어렵다. 본 연구에서는 시간 차이에 따른 추천 효과를 측정하기 위해 실제 온라인 쇼핑몰을 이용하는 고객을 대상으로 개인화된 추천 제품 리스트를 제공하였다. 이를 통해 추천 제품 리스트가 자극이 되어 실제 제품 구매로 연결되는 자극의 효과를 측정하고, 자극의 효과가 발생하는 경우 자극 효과의 지속 시간을 파악하고자 한다. 실험 결과, 온라인 환경에서 실제 고객에게 제품을 추천하는 경우 자극으로 인식되어 고객의 제품 구매율이 증가하는 것을 확인하였으며, 자극의 효과가 감소하는 시점을 파악하였다. 본 연구의 실험 결과를 바탕으로 시간 차이에 따른 자극의 효과를 측정하고 효과의 감소 시점을 파악함으로써 기업에게 추천 시스템 전략 수립에 대한 다양한 시사점을 제공 할 수 있다.