Effect of mobile food environments on fast food visits.

Poor diets are a leading cause of morbidity and mortality. Exposure to low-quality food environments saturated with fast food outlets is hypothesized to negatively impact diet. However, food environment research has predominantly focused on static food environments around home neighborhoods and generated mixed findings. In this work, we leverage population-scale mobility data in the U.S. to examine 62M people's visits to food outlets and evaluate how food choice is influenced by the food environments people are exposed to as they move through their daily routines. We find that a 10% increase in exposure to fast food outlets in mobile environments increases individuals' odds of visitation by 20%. Using our results, we simulate multiple policy strategies for intervening on food environments to reduce fast-food outlet visits. This analysis suggests that optimal interventions are informed by spatial, temporal, and behavioral features and could have 2x to 4x larger effect than traditional interventions focused on home food environments.

Population mobility data provides meaningful indicators of fast food intake and diet-related diseases in diverse populations.

The characteristics of food environments people are exposed to, such as the density of fast food (FF) outlets, can impact their diet and risk for diet-related chronic disease. Previous studies examining the relationship between food environments and nutritional health have produced mixed findings, potentially due to the predominant focus on static food environments around people's homes. As smartphone ownership increases, large-scale data on human mobility (i.e., smartphone geolocations) represents a promising resource for studying dynamic food environments that people have access to and visit as they move throughout their day. This study investigates whether mobility data provides meaningful indicators of diet, measured as FF intake, and diet-related disease, evaluating its usefulness for food environment research. Using a mobility dataset consisting of 14.5 million visits to geolocated food outlets in Los Angeles County (LAC) across a representative sample of 243,644 anonymous and opted-in adult smartphone users in LAC, we construct measures of visits to FF outlets aggregated over users living in neighborhood. We find that the aggregated measures strongly and significantly correspond to self-reported FF intake, obesity, and diabetes in a diverse, representative sample of 8,036 LAC adults included in a population health survey carried out by the LAC Department of Public Health. Visits to FF outlets were a better predictor of individuals' obesity and diabetes than their self-reported FF intake, controlling for other known risks. These findings suggest mobility data represents a valid tool to study people's use of dynamic food environments and links to diet and health.

Predicting merchant future performance using privacy-safe network-based features.

Small and Medium-sized Enterprises play a significant role in most economies by contributing to job creation and economic growth. A majority of such merchants rely on business financing, and thus, financial institutions and investors need to assess their performance before making decisions on business loans. However, current methods of predicting merchants' future performance involve their private internal information, such as revenue and customer base, which cannot be shared without potentially exposing critical information. To address this problem, we first propose a novel approach to predicting merchants' future performance using credit card transaction data. Specifically, we construct a merchant network, regarding customers as bridges between merchants, and extract features from the constructed network structure for prediction purposes. Our study results demonstrate that the performance of machine learning models with features extracted from our proposed network is comparable to those with conventional revenue- and customer-based features, while maintaining higher privacy levels when shared with third-party organizations. Our approach offers a practical solution to privacy concerns over data and information required for merchants' performance prediction, enabling safe data-sharing among financial institutions and investors, helping them make more informed decisions on allocating their financial resources while ensuring that merchants' sensitive information is kept confidential.

A link prediction-based recommendation system using transactional data.

Recommending relevant items to users has become an important task in many systems due to the increased amount of data produced. For this purpose, transaction datasets such as credit card transactions and e-commerce purchase histories can be used in recommendation systems to understand underlying user interests by exploiting user-item interactions, which can be a powerful signal to perform this task. This study proposes a link prediction-based recommendation system combining graph representation learning algorithms and gradient boosting classifiers for transaction datasets. The proposed system generates a network where nodes correspond to users and items, and links represent their interactions. A use case scenario is examined on a credit card transaction dataset as a merchant prediction task that predicts the merchants where users can make purchases in the next month. Performances of common network embedding extraction techniques and classifier models are evaluated via various experiments conducted and based on these evaluations, a novel system is proposed, and a matrix factorization-based alternative recommendation method is compared with the proposed model. The proposed method has shown superior performance to the alternative method in terms of receiver operating characteristic curves, area under the curve, and mean average precision metrics. The use of transactional data for a recommendation system is found to be a powerful approach to making relevant recommendations.

Using gravity model to make store closing decisions: A data driven approach.

Many studies propose methods for finding the best location for new stores and facilities, but few studies address the store closing problem. As a result of the recent COVID-19 pandemic, many companies have been facing financial issues. In this situation, one of the most common solutions to prevent loss is to downsize by closing one or more chain stores. Such decisions are usually made based on single-store performance; therefore, the under-performing stores are subject to closures. This study first proposes a multiplicative variation of the well-known Huff gravity model and introduces a new attractiveness factor to the model. Then a forward-backward approach is used to train the model and predict customer response and revenue loss after the hypothetical closure of a particular store from a chain. In this research the department stores in New York City are studied using large-scale spatial, mobility, and spending datasets. The case study results suggest that the stores recommended being closed under the proposed model may not always match the single store performance, and emphasizes the fact that the performance of a chain is a result of interaction among the stores rather than a simple sum of their performance considered as isolated and independent units. The proposed approach provides managers and decision-makers with new insights into store closing decisions and will likely reduce revenue loss due to store closures.

Validating Gravity-Based Market Share Models Using Large-Scale Transactional Data.

Customer patronage behavior has been widely studied in market share modeling contexts, which is an essential step toward estimating retail sales and finding new store locations in a competitive setting. Existing studies have conducted surveys to estimate merchants' market share and factors of attractiveness to use in various proposed mathematical models. Recent trends in Big Data analysis allow us to better understand human behavior and decision making, potentially leading to location models with more realistic assumptions. In this article, we propose a novel approach for validating the Huff gravity market share model, using a large-scale transactional dataset that describes customer patronage behavior at a regional level. Although the Huff model has been well studied and widely used in the context of sales estimation, competitive facility location, and demand allocation, this article is the first in validating the Huff model with a real dataset. Our approach helps to easily apply the model in different regions and with different merchant categories. Experimental results show that the Huff model fits well when modeling customer shopping behavior for a number of shopping categories, including grocery stores, clothing stores, gas stations, and restaurants. We also conduct regression analysis to show that certain features such as gender diversity and marital status diversity lead to stronger validation of the Huff model. We believe we provide strong evidence, with the help of real-world data, that gravity-based market share models are viable assumptions for retail sales estimation and competitive facility location models.

Quantifying Insurance Agency Channel Dynamics Using Premium Sales Big Data and External Factors.

In insurance business, product sales can be realized over a variety of channels such as independent agencies, or bank branches. In 2017, 55% of premium production was generated over insurance agencies in Turkey making independent agency evaluation prominent in the domain. Unfortunately lacking attention from the scientific community, agency evaluation problem is usually tackled in the industry by utilizing internal business dynamics data. To incorporate the external facts to the agency evaluation process, we propose a computational approach to model behavior traits reflecting insurance agency channel dynamics based on not only premium sales big data but also external facts. We demonstrate how we translate these behavior traits into useful features, namely, utilization, response, and governance, so that each agency can be positioned in a space whose dimensions are determined by these features allowing easy visual detection of segments. Utilization model suggests that each agency has a potential based on its location, determined by several local socioeconomic factors, and it explains the capability of converting potential to profit. To compute utilization scores, we adapt point-of-interest data as a parameter to the segmentation model, a novel approach not only in the insurance business but also in the literature. The response model suggests that a responsive agency must follow overall profit trends of the company. Finally, the governance model explains agency/company cooperation in terms of premium production. All together, we propose a segmentation-based agency evaluation model providing understanding of insurance agency behavior that could be explained and formulated along these three dimensions. Based on the findings from a year-long case study and a proceeding implementation period of our models on an actual analytic system of the insurance company donating the data, we reflect on the performance and usability of our behavioral models that were fit on premium sales big data comprising 127 million transactions. Our results suggest that (1) our approach is quite efficient in extracting features from production logs, (2) behavioral models are quite intuitive resulting in straightforward application steps, and (3) the adoption of behavior models in agency segmentation and evaluation processes is an improvement over commonplace approaches in which premium production is used as the sole metric.

Using Behavioral Analytics to Predict Customer Invoice Payment.

Experiences from various industries show that companies may have problems collecting customer invoice payments. Studies report that almost half of the small- and medium-sized enterprise and business-to-business invoices in the United States and United Kingdom are paid late. In this study, our aim is to understand customer behavior regarding invoice payments, and propose an analytical approach to learning and predicting payment behavior. Our logic can then be embedded into a decision support system where decision makers can make predictions regarding future payments, and take actions as necessary toward the collection of potentially unpaid debt, or adjust their financial plans based on the expected invoice-to-cash amount. In our analysis, we utilize a large data set with more than 1.6 million customers and their invoice and payment history, as well as various actions (e.g., e-mail, short message service, phone call) performed by the invoice-issuing company toward customers to encourage payment. We use supervised and unsupervised learning techniques to help predict whether a customer will pay the invoice or outstanding balance by the next due date based on the actions generated by the company and the customer's response. We propose a novel behavioral scoring model used as an input variable to our predictive models. Among the three machine learning approaches tested, we report the results of logistic regression that provides up to 97% accuracy with or without preclustering of customers. Such a model has a high potential to help decision makers in generating actions that contribute to the financial stability of the company in terms of cash flow management and avoiding unnecessary corporate lines of credit.

Customer mobility signatures and financial indicators as predictors in product recommendation.

The rapid growth of mobile payment and geo-aware systems as well as the resulting emergence of Big Data present opportunities to explore individual consuming patterns across space and time. Here we analyze a one-year transaction dataset of a leading commercial bank to understand to what extent customer mobility behavior and financial indicators can predict the use of a target product, namely the Individual Consumer Loan product. After data preprocessing, we generate 13 datasets covering different time intervals and feature groups, and test combinations of 3 feature selection methods and 10 classification algorithms to determine, for each dataset, the best feature selection method and the most influential features, and the best classification algorithm. We observe the importance of spatio-temporal mobility features and financial features, in addition to demography, in predicting the use of this exemplary product with high accuracy (AUC = 0.942). Finally, we analyze the classification results and report on most interesting customer characteristics and product usage implications. Our findings can be used to potentially increase the success rates of product recommendation systems.

Money Walks: Implicit Mobility Behavior and Financial Well-Being.

Traditional financial decision systems (e.g. credit) had to rely on explicit individual traits like age, gender, job type, and marital status, while being oblivious to spatio-temporal mobility or the habits of the individual involved. Emerging trends in geo-aware and mobile payment systems, and the resulting "big data," present an opportunity to study human consumption patterns across space and time. Taking inspiration from animal behavior studies that have reported significant interconnections between animal spatio-temporal "foraging" behavior and their life outcomes, we analyzed a corpus of hundreds of thousands of human economic transactions and found that financial outcomes for individuals are intricately linked with their spatio-temporal traits like exploration, engagement, and elasticity. Such features yield models that are 30% to 49% better at predicting future financial difficulties than the comparable demographic models.