Search

Studies have found that a high level of work engagement links to better business performance, higher revenue growth, higher profitability and higher customer satisfaction, yet according to Gallup, only 13% of global employees are engaged at work. Leadership has been found to play a central role in improving employee work engagement level, yet the mechanism has yet to be fully understood. This research explores the effects of leadership styles (servant leadership and abusive leadership) on employee work engagement, taking into account the mediating role of psychological safety and the moderator role of leader distance. The data used in this study is a subset of the data collected through a broader research project run by Exponential team. The questionnaires were sent to 4046 participants in Finnish companies, which resulted in 1130 nearly complete responses. The respondents are employees working in eleven Finnish SMEs and corporations. Confirmatory factor analysis and structural equation modeling were used in analyzing the data. The findings suggest that psychological safety partially mediates the relationship between servant leadership and work engagement and fully mediates the relationship between abusive leadership and work engagement. As predicted, servant leadership is positively related to work engagement and abusive leadership is negatively related to work engagement. Leader distance is found to negatively moderates the relationship between abusive leadership and psychological safety while having no effects on the relationship between servant leadership and psychological safety. The study provides insights into how employee’s psychological well-being can affect employee engagement and how leader behaviors can promote or discourage employees from engaging at work. By including leader distance as a moderator, the study also considers how leadership influences are affected by today’s new work practices: remote work and virtual work. Overall, the study findings ad

Electronic commerce industry has emerged just in decades, yet prospered tremendously and totally shaped the global economy with various newly introduced digital products and services far from our ever imaginations. Along with online shopping progresses, fraudulent activities have also evolved continually and hugely damaged both profit of businesses and money of customers. Thus, fraud detection has continuously been a hot topic with the exponentially growing concerns from normal practitioners to specialized researchers. The fight against frauds requires huge efforts with time and associated costs since several main challenges significantly hindered the concerned researches as well as the prediction performances of modeling techniques. These constraints involve the highly imbalanced issue in datasets, the concept drift in customers and fraudsters, the delayed supervised information and other real-world problems. Following that, the most state-of-art solutions dealt with these challenges are demonstrated systematically in three levels, involving data-level approaches, algorithm-level techniques and assessment-level methods. Based on the above solutions, the thesis conducts the empirical study on an ecommerce dataset to examine the classification performances of different combined techniques in capturing efficiently online frauds. Accordingly, a methodology framework is proposed, employing four distinctive resampling methods, involving ROS, RUS, SMOTE and SmoteTomek, coupled with three learning algorithms Logistic Regression, Random Forest and eXtreme Gradient Boosting for the model experimentation part. The research has asserted that the combined approaches of different techniques would tackle well the online fraud detection, promisingly relieving many pressures for online merchants. Additionally, the findings have also reaffirmed the validity of resampling techniques and provided an insight into appropriate assessment metrics employed under the imbalanced scenarios. O

This study focuses on the order picking process of a distribution center (DC) supplying locomotive and railroad car parts. The DC employs a manual picker-to-parts system where pickers move on foot or by vehicular means. Efficiency of an order picking process in such a DC mainly depends on three problems when the layout of the DC is given: the storage location assignment problem (SLAP), the order batching problem and the order picker routing problem. This study focuses on the aggregate effect three major decisions have on order picking performance in a manual picker-to-parts warehouse. To study the effects of these three sub-problems, we create a framework that allows us to run simulated scenarios with different approaches to these problems. To solve the SLAP, we employ a hybrid of class-based and family grouping methods by enhancing a class-based SKU location assignment with modern clustering techniques. To further improve the order picking process, we experiment with various order batching methods. We use picker routing heuristics to evaluate combinations of the storage location assignments and batching procedures. Over a set of order lines to be fulfilled, the objective function is be the aggregate distance covered over the warehouse floor. We show that distance savings of more than 55% can be achieved by rearranging the DC. Moreover, we show that stock keeping unit (SKU) clustering can improve the performance of class-based storage location assignments and that even simple order batching algorithms are likely to improve order picking performance significantly. Based on the framework, we develop a tool set that encompasses the aspects concerning the DC’s order picking process. The solution will be implemented into a cloud-computing environment, allowing for real-time tracking of the DC’s order picking efficiency and the generation of visual tools that help move SKUs to desirable shelf locations and batch orders.

Mobile traffic and number of connected devices have been increasing exponentially nowadays, with customer expectation from mobile operators in term of quality and reliability is higher and higher. This places pressure on operators to invest as well as to operate their growing infrastructures. As such, telecom network management becomes an essential problem. To reduce cost and maintain network performance, operators need to bring more automation and intelligence into their management system. Self-Organizing Networks function (SON) is an automation technology aiming to maximize performance in mobility networks by bringing autonomous adaptability and reducing human intervention in network management and operations. Three main areas of SON include self-configuration (auto-configuration when new element enter the network), self-optimization (optimization of the network parameters during operation) and self-healing (maintenance). The main purpose of the thesis is to illustrate how anomaly detection methods can be applied to SON functions, in particularly self-healing functions such as fault detection and cell outage management. The thesis is illustrated by a case study, in which the anomalies - in this case, the failure alarms, are predicted in advance using performance measurement data (PM data) collected from a real LTE network within a certain timeframe. Failures prediction or anomalies detection can help reduce cost and maintenance time in mobile network base stations. The author aims to answer the research questions: what anomaly detection models could detect the anomalies in advance, and what type of anomalies can be well-detected using those models. Using cross-validation, the thesis shows that random forest method is the best performing model out of the chosen ones, with F1-score of 0.58, 0.96 and 0.52 for the anomalies: Failure in Optical Interface, Temperature alarm, and VSWR minor alarm respectively. Those are also the anomalies can be well-detected by the mode

The sports betting market has grown quickly over the past the few decades mostly due to the digitalization of the business. The bookmakers have moved online from the physical locations. The market has thus globalized, and the competition has increased, forcing the bookmakers to produce more accurate estimates about the events to keep up with the competition. This study investigates the accuracy of NHL moneyline betting odds, in predicting the actual outcomes of games throughout the time that the betting events are open. The dataset covers three full seasons from 2015 to 2018. The odds are collected at 5 different time points for each game and the differences in the predictive power of time points is analyzed. Then, the odds and their movement is investigated further to see if there are profitable betting strategies to be found based solely on information about odds movement. The tests related to the prediction accuracy of the odds reveal that there are no statistically significant differences between the prediction accuracy for different time points. Aggregate results are rather consistent though in showing that before the day of the game, the estimates implied by the odds aren’t quite as accurate as they are during the game day. The regression tests further indicate that when the implied probability of a selection grows, the objective probability grows at a higher rate, meaning that there’s a Favorite-Longshot Bias in the market. This means that betting on a more likely outcome yields better returns on average. The tests for finding profitable betting strategies further enforce the notion of Favorite-Longshot Bias and subsequently all of the consistently profitable betting strategies, that are found, involve only betting on favorites. The data about the odds movement between time points and splitting the teams to favorites and underdogs reveals that betting on favorite teams who have had their odds rising in a given time point interval, yields a profit for 80% of t

[EN] Customer retention has been the focus of customer relationship management in the financial sector during the past decade. The first and important step in customer retention is to classify the customers into possible churners, those likely to switch to another service provider, and non-churners. The second step is to take action to retain the most probable churners. The main challenge in churn classification is the rarity of churn events. In order to overcome this, two aspects are found to improve the churn classification model: the training data and the algorithm. The recently proposed multi-period training data approach is found to outperform the single period training data thanks to the more effective use of longitudinal data. Regarding the churn classification algorithms, the most advanced and widely employed is the ensemble method, which combines multiple models to produce a more powerful one. Two popularly used ensemble techniques, random forest and gradient boosting, are found to outperform logistic regression and decision tree in classifying churners from non-churners. The study uses data of housing loan customers from a Nordic bank. The key finding is that models combining the multi-period training data approach with ensemble methods performs the best.

A car is a relatively expensive and long-term purchase for a consumer. Its operating costs can be expected to play a significant role in purchase decisions. Prices of gasoline and diesel fuels vary over time, and this makes diesel powered vehicles relatively more or less cost-efficient to operate depending on the price situation of fuels. In this thesis, it is investigated if a long run level relationship exists between consumer prices of gasoline and diesel, and proportions of diesels in three popular car models registered in Finland between 2005 and 2015. Also, the existing literature on car choice is reviewed to justify the expectation of such a relationship, and to identify any prior evidence of it. Existing models of car choice are commonly based on random utility theory, and employ discrete choice models in describing and predicting consumers’ car choices among a discrete set of alternatives. While operating costs are commonly included in car choice models, fuel prices themselves are more seldom considered. However, changing fuel prices may make a diesel model more cost-efficient to run if diesel is relatively cheap, which would theoretically lead to a bigger diesel share of cars purchased, assuming the decision maker acts rationally. An Autoregressive Distributed Lag with Bounds Testing method is used in this thesis to investigate if a long run level relationship between prices of gasoline and diesel, and diesel shares in three popular cars exists. The method is preferable to the type of setting in this thesis, where the time series may be stationary, non-stationary or cointegrated in the same equation. The car models chosen into the analysis are VW Golf, Toyota Avensis and Skoda Octavia, because these have been among the most popular cars during the focus period, and therefore a sufficient amount of data exists. It is found that in two out of three cars analyzed, the long run level relationship between fuel prices and diesel shares exists. Further, the direc

Within the field of statistical analysis in sports and football, studying teams’ playing style has not been a popular domain. The quantification of football styles based on gameplay data proposed in this research provides a statistical approach to analyze the underlying football styles adopted by different teams during a competitive tournament. The main empirical design of this research lies in the application of exploratory factor analysis to a set of performance-related variables; the objective is to reveal underlying football styles of the teams through the latent factors. Factoring is a practical and helpful method that can be used for investigating variable relationships for complex concepts, hopefully to reveal underlying patterns in a set of variables. Factor analysis is employed by researchers to examine concepts that are not directly measured through contracting a big number of variables into fewer underlying factors that allow better interpretation. The FIFA World Cup is a highly interesting source to study different styles of football. By analysing an extensive data set collected for the 32 teams and more than 600 players competing during the 2014 FIFA World Cup in Brazil, the results from this study assert that there are three underlying playing styles among the pool of teams participating the World Cup that can be derived from the data set: possession-based, individualistic, and practical style. An extended analysis provides further insights into the impact that background attributes of the players in the team can have on the underlying styles. The results stemming from the study is expected to contribute to the literature by providing a fresh approach in studying the styles of play adopted by football teams based on the use of exploratory factor analysis. As the players are a critical part of overall team performance, which eventually impacts the financial and business performance of a sporting team, decision makers can utilize the approach presented i

Customer retention has been the focus of customer relationship management research in the financial sector during the past decade. The first step in customer retention is to classify the customers into binary groups of possible churners, meaning customers that are likely to switch to another service provider, and non-churners, referring to those that are probably staying with the current provider. The second step in customer retention is to take action to retain the most probable churners to either minimize costs or maximize benefits. As a result, churn classification is an important first step in customer retention. However, the main challenge in churn classification is the extreme rarity of churn events. For example, the churn rate in the banking industry is usually less than 1%. In order to overcome this rarity issue, a great deal of research has been found to improve the two main aspects of a churn classification model: the training data and the algorithm. Regarding the training data, the recently proposed multi-period training data approach is found to outperform the single period training data thanks to the more effective use of longitudinal data of churn behavior. Regarding the churn classification algorithms, the most advanced and widely employed is ensemble method, which combines multiple models to produce a more powerful one. Two popularly used ensemble techniques are random forest and gradient boosting, both of which are found to outperform logistic regression and decision tree in classifying churners from non-churners. To the best of the author’s knowledge, the proposed multi-period training data has not been applied to the ensemble methods in a churn classification model. As a result, the thesis would like to study whether this multi-period training data approach, when employed together with ensemble methods in a churn classification model, produces better churn prediction than with logistic regression and decision tree. The ensemble methods used in thi