We share our empirical results in "Deep Learning in Asset Pricing" by providing the access to our two asset pricing models for individual stock returns. One model generates SDF weights, and the other model predicts beta. These two models produce a panel of SDF weights and beta's automatically using your own data. Instead of putting in all macro variables, you could simply put in a historical time t and the characteristics associated with each stock, and we would give the weights and beta's given the values of the macro variables at that time.
The input csv file should ontain 49 columns with full observations: date, permno, return (ret) and 46 stock-level characteristics. sample.csv is a sample input file.
The output csv file contains 4 columns: date, permno, SDF weight (w) and beta. sample_out.csv is a sample output file.
We take both raw and processed firm characteristics as inputs. However, we do recommend that you process your data and normalize characteristics between [-0.5,0.5]. Please check the box "Processed" if you have processed your data.
If successful, the output file will be downloaded automatically within 30 seconds. Otherwise, you will be redirected to a page showing "status: 500". The router will terminate the request if it takes longer than 30 seconds to complete. Please limit the size of your input files to 10 MB.