2_code

Does Machine Learning outperform Logistic Regression in predicting individual tree mortality?

💻 💾 📊 Original data, code and results related to the study

---

💡🧠 Each file and/or folder code corresponds to the script used to generate it, ensuring that all (dataset + script + output) share the same code

⚠️ 📜 Remember to update the script paths in your working directory if you plan to use that code

---

📁 Folder Content

• 📜 0_data_curation.r

  • 💡 Purpose: Manage initial data: structure, IDs, input missing data.
  • 💾 ➡️ 💻 Input: 1_raw/final/VF_daten.xlsx, 1_raw/final/Fi-Daten__age.xlsx
  • 💻 ➡️ 💾 Output: 1_processed/0_initial_df_clean/initial_df_clean.csv

• 📜 1.0_neighborhood_main.r, 1.1_neighborhood_functions.r

  • 💡 Purpose: Calculate variables needed for the analysis using a subplot of 0.33*h radii around each tree.
  • 💾 ➡️ 💻 Input: 1_data/1_processed/0_initial_df_clean/initial_df_clean.csv
  • 💻 ➡️ 💾 Output: 1_data/1_processed/1_neighborhood/*, trees_r33.csv, subplot_stats_r33.csv, neighborhood_stats_r33.csv

• 📜 2_climate_data.r

  • 💡 Purpose: Calculate climate variables by plot location. ⚠️ Requires prior download of climate data from WorldClim, contact authors for details if needed.
  • 💾 ➡️ 💻 Input: 1_raw/final/Koordinaten.xlsx
  • 💻 ➡️ 💾 Output: 1_data/1_processed/2_clima/df_complete_r33.csv

• 📜 3_feature_visualization.r

  • 💡 Purpose: Generate graphs and manually study variable relationships.
  • 💾 ➡️ 💻 Input: 1_data/1_processed/2_clima/df_complete_r33.csv
  • 💻 ➡️ 💾 Output: None

• 📜 4.0_split_dataset.r, 4.1_split_variables.r, 4.2_functions_var_combis.r

  • 💡 Purpose: Split datasets (size and thinning) and variables for case studies.
  • 💾 ➡️ 💻 Input: 1_data/1_processed/2_clima/df_complete_r33.csv
  • 💻 ➡️ 💾 Output: 1_data/1_processed/4_datasets/*

• 📜 5.0_run_analysis.r, 5.1_LR_analysis.r, 5.2_DT_analysis.r, 5.3_RF_analysis.r, 5.4_NB_analysis.r, 5.5_KNN_analysis.r, 5.6_SVM_analysis.r

  • 💡 Purpose: Run analysis (except ANN) in R for different case studies.
  • 💾 ➡️ 💻 Input: 1_data/1_processed/4_datasets/*
  • 💻 ➡️ 💾 Output: 1_data/1_processed/5_analysis/**case_study**/*, metrics.RData, models.RData

• 📜 6_HPC

  • 💡 Purpose: Run simulations on iuFOR HPC, split by study case.
  • 💾 ➡️ 💻 Input: 1_data/1_processed/4_datasets/*
  • 💻 ➡️ 💾 Output: 1_data/1_processed/5_analysis/**case_study**/*, metrics.RData, models.RData

• 📜 7_metrics_compilation.r

  • 💡 Purpose: Extract R analysis metrics and create a checkpoint.
  • 💾 ➡️ 💻 Input: 1_data/1_processed/5_analysis/*, **case_study**/metrics.RData, ann/preds/**case_study**/*, ann/timer/**case_study**/*
  • 💻 ➡️ 💾 Output: 1_data/1_processed/6_final_results/**case_study**/final_metrics.RData

• 📜 8.0_performance_graphs.r, 8.1_functions_performance_graphs.r, 8.2_classifiers_comparison.r, 8.3_graph_functions.r, 8.4_application_thinning.r, 8.5_application_thinning_comparison.r, 8.6_time_and_performance_graphs.r

  • 💡 Purpose: Compare analysis metrics across case studies using graphs. Graphs in the original paper use 8.5 and 8.6.
  • 💾 ➡️ 💻 Input: 1_data/1_processed/6_final_results/**case_study**/final_metrics.RData
  • 💻 ➡️ 💾 Output: 2_scripts/4_figures/*

• 📜 9.0_location_map.r, 9.1_neighbour_graphs.r, 9.2_mortality_graphs.r, 9.3_df_mortality_rates.r, 9.4_paper_tables.r

  • 💡 Purpose: Generate graphs (location, neighborhood) and obtain tables for the original paper.
  • 💾 ➡️ 💻 Input: 1_raw/final/Koordinaten.xlsx, 1_data/1_processed/2_clima/df_complete_r33.csv, 1_data/1_processed/1_neighborhood/trees_r33.csv
  • 💻 ➡️ 💾 Output: 2_scripts/4_figures/*

---

📚 Additional Information

A flowchart detailing the training and testing process (scripts from groups 5 and 6) is shown here: