background

In situations with inflated prices, such as housing costs, we can only make qualitative predictions. It's not possible to quantitatively determine how long the price increase will continue or how much it will rise.

3. 서울 집값 예측

• 2019년 당시 상승중이던 지가지수 데이터로부터 2022년 최고점과 상승율, 그리고 최저 하락 지점과 하락율 예측

• 2019 당시 데이터로부터 LSTM과 RNN 등과 같은 데이터를 시간적으로 후행하는 모델은 사용할 수 없는 판단

• 미국, 일본, 스페인 등 여러 나라의 데이터를 사용

• ANN 모델을 직접 구현하여 파라미터가 5개인 분포 모델(분포의 크기를 나타내는 파라미터 제외하고 분포를 고정 가능하기 때문)을 가지고 데이터 피팅

• 서울 지가지수 데이터에 맞게 파라미터 튜닝

• 2022년 최고점의 시점과 상승율 그리고 현재까지 오차범위 내 적중

abstract

"This project involves designing a quantitative model by integrating Hyman Minsky's economic model with Planck's law describing the spectral density of blackbody. This model enables the prediction of prices, including those influenced by bubbles, and allows for the assessment of the presence of bubbles in prices.”

prediction result

• "This project, conducted in the summer of 2019, involved testing South Korean Seoul housing price data at that time. It predicted that the prices would rise up to 1.5 times their value at that time and then start to decline by 2022 (the exact month was not calculated)." -> Looking at it in September 2023, it turned out to be exactly correct.

Key Achievements

1. As a quantitative model based on Hyman Minsky's qualitative bubble economy model, it provides a numerical description of Minsky's economic model.
2. Additionally, it utilizes an indicator denoted as 'T' representing temperature to provide a single metric for quantifying the heat of an overheated market.
3. As a result, the model is fitted with abnormally formed price data, allowing it to assess and predict price increase periods, peak prices, and the extent of bubble inclusion.