In this module there are three different data science / analysis projects. The first is concerned with network analysis, the second with Kalman and Extended Kalman Filters and the last is the model developement to estimate housing price data.
The document analyzes a network of corporations, based on Twitter's publicly available data. First, a machine learning models is trained to predict whether or not an account belongs to a corporation. This model is then used to estimate whether or not user accounts belong to corporations or not. Second, five most influential corporations are being identified. Third, the network is divided into homogenous communities by comparing community detection algorithms. Fourth, the two previous results are combined to identify the most influential user in each community. Finally, associations of corporations and interest groups are discovered. There are two main associations: either corporations are affiliated to hashtags relating to innovation and technology or two holidays.
This document provides an overview of Kalman-Filter, Extended-Kalman-Filter and Particle-Filter. These are prominent examples of a class of estimators called bayesian filters. This paper uses the properties of the multivariate gaussian distribition to derive key results for the Kalman-Filter, and provides examples of applications of state-space models.
The aim of this work is to obtain a model for estimating the net rental price. For this purpose a extensive data control and analysis carried out. Using the best subset method, the best data based linear model was estimated. Add interaction terms failed to lead to significant model improvements. The model has the following problems: On the one hand, there are inhomogeneous variances. For this model led attempting to fix this problem but with worse results. On the other hand, data contains outliers. Due to a lack of objective justification, the outliers will remain in the model. For these reasons, a linear model with all provided variables is used to estimate the net rental price.