2021/05/11 by Jan-Peter Kucklick, Oliver Müller, Kucklick, Jan-Peter +1
Environmental Science · Economics, Econometrics and Finance · #Urban Planning and Valuation #Housing Market and Economics
paper · pdf · doi:10.48550/arxiv.2105.04984
In the house credit process, banks and lenders rely on a fast and accurate\nestimation of a real estate price to determine the maximum loan value. Real\nestate appraisal is often based on relational data, capturing the hard facts of\nthe property. Yet, models benefit strongly from including image data, capturing\nadditional soft factors. The combination of the different data types requires a\nmulti-view learning method. Therefore, the question arises which strengths and\nweaknesses different multi-view learning strategies have. In our study, we test\nmulti-kernel learning, multi-view concatenation and multi-view neural networks\non real estate data and satellite images from Asheville, NC. Our results\nsuggest that multi-view learning increases the predictive performance up to 13%\nin MAE. Multi-view neural networks perform best, however result in\nintransparent black-box models. For users seeking interpretability, hybrid\nmulti-view neural networks or a boosting strategy are a suitable alternative.\n