Browsing by Author "Ervasti, Mikko"

Filter results by typing the first few letters
Now showing 1 - 9 of 9
  • Adapting marketing mix modelling for the retail marketing environment – A road map for development
    (2019-10-23) Heliste, Antti
     Perustieteiden korkeakoulu | Master's thesis
    Measuring the impact of marketing is essential for improving its performance and justifying marketing decisions to top management. However, marketers often struggle with it, even though various methods are available to them in the literature. A good starting point for marketers is the most popular method, marketing mix modelling (MMM), that is a linear regression fitted in sales and marketing data. Yet, it often suffers from various downsides, such as lack of data, deficient model forms and biases. Researchers have consequently suggested improving it through better data, better models and model validation. However, researchers mainly discuss these areas as a way to improve model accuracy rather than to widen the scope of analysis. Improving modelling granularity would enable marketers to analyse performance on lower levels and broaden their discussion on improvements. Higher granularity could particularly support the retail industry, where marketing is a complex operation because of wide product ranges, geographical reaches and customer bases. Consequently, the goal of the thesis was to analyse how the typical MMM is limited, how model developers could adjust it to meet the needs of the retail marketing environment and what impacts such adjustments would have. We conducted the study through a combination of a literature review and a simulation. The literature review discovered that the typical MMM is limited in use in the retail environment, mainly due to its low granularity that hides the information about the structure of performance. Other flaws include, e.g., the lack of modelling in retail-specific effects, such as stock-up, and the lack of model validation. The most significant opportunity arises from increasing granularity in at least three dimensions: frequency, geography and product hierarchy. Other improvements, in turn, arise from improving accuracy through comprehensive modelling and model validation through simulation. The simulation studied the impact of granularity on available improvement opportunities in the retail environment. A product-level model was able to reach a significant 33.1\% increase in total profit compared to the unoptimised baseline. The traditional model, in turn, was only able to reach a meagre 1.7\% improvement. The result supports the hypothesis that higher modelling granularity leads to more detailed and effective improvement opportunities in retail marketing. Based on the literature review and the simulation, we formed a road map for the development of MMM in the retail environment.
  • Bayesian model validation and selection metrics in retail time series forecasting
    (2021-09-26) Rönty, Leevi
     Perustieteiden korkeakoulu | Bachelor's thesis
  • Evaluating cannibalization between items in retail promotions
    (2018-10-16) Herrala, Olli
     Perustieteiden korkeakoulu | Bachelor's thesis
  • Quality of analytics management of data pipelines for retail forecasting
    (2019-08-19) Kreics, Krists
     Perustieteiden korkeakoulu | Master's thesis
    This thesis presents a framework for managing quality of analytics in data pipelines. The main research question of this thesis is the trade-off management between cost, time and data quality in retail forcasting. Generally this trade-off in data analytics is defined as quality of analytics. The challenge is addressed by introducing a proof of concept framework that collects real time metrics about the data quality, resource consumption and other relevant metrics from tasks within a data pipeline. The data pipelines within the framework are developed using Apache Airflow that orchestrates Dockerized tasks. Different metrics of each task are monitored and stored to ElasticSearch. Cross-task communication is enabled by using an event driven architecture that utilizes a RabbitMQ as the message queue and custom consumer images written in python. With the help of these consumers the system can control the result with respect to quality of analytics. Empirical testing of the final system with retail datasets showed that this approach can aid data science teams to provide better services on demand with bounded resources especially when dealing with big data.
  • Quantum Entanglement Properties of Few-Qubit Systems
    (2010) Ervasti, Mikko
     Informaatio- ja luonnontieteiden tiedekunta | Bachelor's thesis
  • Simulating impurities and edges in graphene
    (2016) Ervasti, Mikko
     School of Science | Doctoral dissertation (article-based)
    Graphene is a two-dimensional allotrope of carbon with incredible mechanical strength, high charge carrier mobility and excellent thermal conductivity. These remarkable properties present numerous potential applications in nanoelectronics and related fields. However, using graphene in a field-effect transistor requires opening a band gap, which can be achieved by cutting graphene into ribbons. Furthermore, the electronic structure and transport properties of graphene are modified by various kinds of defects, such as vacancies, impurities and grain boundaries. Both the defects and edges can host magnetic states that are useful in spintronics applications. In this Thesis, impurities and edges in graphene are simulated using computational techniques. Part of the research has been done in collaboration with experimental groups. The computational simulations provide the necessary link between theory and experiment, aiding in the interpretation of the measurements. The main computational methods used are tight-binding, exact diagonalization and density functional theory, of which the tight-binding and exact diagonalization methods were implemented by the author. Exact diagonalization was used to evaluate correlation energies and reference data to exchange-correlation functionals in two-dimensional quantum dots, electron-positron annihilation in three-dimensional quantum dots, and many-body properties of finite graphene nanoribbons. The research sheds light on the electronic and magnetic properties of graphene. By using the first-principles density functional theory, the formation energies of silicon and silicon-nitrogen impurities were evaluated to identify the relevant low-energy configurations. By fitting to tight-binding models, the transport properties of systems containing randomly distributed impurities were determined. Moreover, hydrogen adatoms with noncollinear spins were shown to scatter the electron spin strongly close to the charge neutrality point. The narrow finite graphene nanoribbons were found to have only small band gaps, and the simulated scanning tunneling microscopy maps and spectra of the ribbons agreed with the experiments. The precise atomic structure at the graphene-hexagonal boron nitride interfaces was determined with the help of simulations, and the interfaces were shown to host electronic states similar to those on the graphene edges. Overall, the theoretical and computational results build up the knowledge and understanding of graphene-related systems.
  • Synthesis of Extended Atomically Perfect Zigzag Graphene - Boron Nitride Interfaces
    (2015) Drost, R.; Shawulienu, Kezilebieke; Ervasti, Mikko; Hämäläinen, Sampsa; Schulz, F.; Harju, A.; Liljeroth, P.
     A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
    The combination of several materials into heterostructures is a powerful method for controlling material properties. The integration of graphene (G) with hexagonal boron nitride (BN) in particular has been heralded as a way to engineer the graphene band structure and implement spin- and valleytronics in 2D materials. Despite recent efforts, fabrication methods for well-defined G-BN structures on a large scale are still lacking. We report on a new method for producing atomically well-defined G-BN structures on an unprecedented length scale by exploiting the interaction of G and BN edges with a Ni(111) surface as well as each other.
  • Topological phases of matter and the fractional quantum Hall effect
    (2011) Ervasti, Mikko
     School of Science | Master's thesis
    The subject of this work is to study the topological phases of matter in (2+1)-dimensions theoretically and numerically. As the theory is not yet complete, we use various gapped model systems to probe the signature behaviour that arise in the topological phase. Most notably, we study the fractional quantum Hall systems, since they are well-established to exist in nature. The main motivation is that the topological phases can be used for topological quantum computation. The first part of the Thesis studies the topological phase from theoretical principles. This consists of defining modular tensor categories to describe consistent quasiparticle systems with arbitrary number of quasiparticles, and topological quantum field theories to describe the low energy, long wavelength effective field theories for the topological phases. There are also other viewpoints that are covered, namely ground state degeneracy that is due to topological properties and not by any symmetry-breaking, and also topological phases forming universality classes of different longranged entanglements. We solve a fractional quantum Hall system on a sphere with Coulomb interaction numerically and evaluate the topological entanglement entropy. This amounts to finding the ground states with various system parameters by an exact diagonalization technique, evaluating the entanglement entropies based on bipartite splits, and extrapolating the entropies to the thermodynamic limit. The subleading term in the entropy then matches the logarithm of the total quantum dimension, which characterizes the quasiparticles. The results match the theoretical predictions.
  • Ultra-narrow metallic armchair graphene nanoribbons
    (2015) Kimouche, A.; Ervasti, Mikko; Drost, R.; Halonen, S.; Harju, A.; Joensuu, Pekka; Sainio, J.; Liljeroth, P.
     A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
    Graphene nanoribbons (GNRs)—narrow stripes of graphene—have emerged as promising building blocks for nanoelectronic devices. Recent advances in bottom-up synthesis have allowed production of atomically well-defined armchair GNRs with different widths and doping. While all experimentally studied GNRs have exhibited wide bandgaps, theory predicts that every third armchair GNR (widths of N=3m+2, where m is an integer) should be nearly metallic with a very small bandgap. Here, we synthesize the narrowest possible GNR belonging to this family (five carbon atoms wide, N=5). We study the evolution of the electronic bandgap and orbital structure of GNR segments as a function of their length using low-temperature scanning tunnelling microscopy and density-functional theory calculations. Already GNRs with lengths of 5 nm reach almost metallic behaviour with ~100 meV bandgap. Finally, we show that defects (kinks) in the GNRs do not strongly modify their electronic structure.