Figure 2
The flowchart consists of a vertical flow of information starting with a rectangular box at the top labeled “Full dataset - 24 features” which contains “(Clarkson’s research and investing.com databases)”. A downward arrow leads to a rectangular box labeled “Section 3.2” which contains “Data collection and preparation”. A downward arrow leads to a capsule-shaped box labeled “Section 3.3” which contains “Structure of D L and A R I M A X models”. From this capsule, two downward arrows lead to different paths. The left path leads to a rectangular box with rounded corners labeled “A R I M A X (1, 1, 1)”. The right path leads to a vertical stack of four rectangular boxes with rounded corners. The first box is labeled “Holdout without D R T s”, which contains “(24 features)”. The second box is labeled “Holdout with the M S E filter” which contains “(6 features)”. The third box is labeled “Holdout with the M S E filter and P C A” which contains “(3 features)”. The fourth box is labeled “Model training, validation and test”. A downward arrow connects each box in this right stack sequentially. A side arrow labeled “feedback” originates from the “Model training, validation and test” box and loops back to the “Holdout without D R T s” box. Finally, a downward arrow from the “A R I M A X (1,1,1)” box and a downward arrow from the “Model training, validation and test” box both lead to a diamond-shaped box at the bottom labeled “Section 3.4, 3.5” which contains “Model evaluation”.Model development flowchart. Source: Authors