Reproducibility Package — ANF-CAF Results and Code (D1–D4)

Reproducibility Package — ANF-CAF Results and Code (D1–D4) This repository provides the companion artifacts for the paper “Robust Agricultural IoT Data Acquisition using Dual Adaptive Noise and Context-Aware Anomaly Filtering.” It contains per-dataset and combined results for four agricultural IoT datasets (greenhouse and open-field signals), benchmarking ANF-CAF against Kalman, regression/trend-preserving filters, and an AI-based denoiser. Artifacts are organized to enable fair, repeatable comparison and quick figure/table regeneration. Contents /data/ results_D1.csv, results_D2.csv, results_D3.csv, results_D4.csv — per-dataset metrics and edge footprints. results_all_datasets.csv — aggregated results across D1–D4 for cross-dataset analysis and summary figures. /tables/ Table_2_CAF_Event_Level_Metrics.csv — precision/recall/F1 and repair outcomes (spikes, drifts, out-of-range), over-correction, time-to-repair. Table_3_Edge_Feasibility_and_Footprint.csv — latency, throughput, memory, energy, and sampling headroom. Table_4_Benchmarking_Across_Datasets.csv — best-method per dataset, margins vs. next best, and ANF-CAF vs. AI deltas. Table_5_Signals_to_Operations.csv — metric-to-operations mapping (actuation risk proxy, repair delay, schedule-jitter proxy, false-alert reduction). Data schema (applies to results_*.csv) DatasetID, DatasetName, Method, RMSE_Raw, RMSE_Filtered, RMSE_ReductionPct, VarianceReductionPct,Anom_PrecisionPct, Anom_RecallPct, Anom_F1Pct, Latency_ms_per_sample, Energy_mJ_per_sample,Memory_kB, Throughput_Hz, N_Records. How to use Recreate tables/figures: Load results_all_datasets.csv and the CSVs in /tables to plot RMSE/variance reductions, F1 scores, and edge budgets. Drill down by dataset: Use results_D*.csv for per-dataset charts or ablation summaries. Report integration: Tables are column-aligned with the manuscript’s results section, enabling direct import into LaTeX or spreadsheet templates. What this enables Fair comparison: Same schema and metrics across all methods/datasets. Operational emphasis: Includes latency/energy/memory for ESP32/NodeMCU feasibility. Adoption: Ready-to-use presets and labeled results support rapid replication and extension.