Fig 1.
Real estate index and stock data preprocessing.
Fig 2.
Basic flow of CNN applied to real estate index and stock.
Fig 3.
Comparison of RNN and LSTM structures (a. RNN; b. LSTM).
Fig 4.
Schematic diagram of LSTM network applied to real estate index and stock prediction.
Fig 5.
The prediction model framework of real estate index and stock trend based on CNN-LSTM algorithm optimized by DL.
Table 1.
Parameter configuration of the CNN, LSTM, and CNN-LSTM algorithm.
Fig 6.
The result of predicting the trend of real estate index and stock based on the CNN algorithm compared with the actual value (a. China Real Estate Index (399241); b. Poly Development (600048); c. Gemdale Group (600383); d. Nanshan Holdings (002314).
Fig 7.
The comparison of the trend of real estate index and stock based on the LSTM algorithm with the actual value (a. China Real Estate Index (399,241); b. Poly Development (600048); c. Gemdale Group (600383); d. Nanshan Holdings (002314).
Fig 8.
The proposed CNN-LSTM algorithm predicting the trend of real estate index and stock (a. China Real Estate Index (399,241); b. Poly Development (600048); c. Gemdale Group (600383); d. Nanshan Holdings (002314).
Table 2.
RMSE and its average values under CNN, LSTM and CNN-LSTM algorithms.
Table 3.
The descriptive statistics of the CNN, LSTM, and CNN-LSTM algorithms.
Table 4.
ANOVA test results of the proposed CNN-LSTM, CNN, and LSTM algorithms.
Table 5.
Descriptive one-sample t test results of the CNN, LSTM, and proposed CNN-LSTM algorithms.
Fig 9.
Influence curve of stock prediction accuracy with increased iterations under different algorithms.
Fig 10.
The influence curve of the time required for stock prediction with the increase of iteration times under different algorithms.