Create snakefile and config

Snakefile

@@ -0,0 +1,103 @@
+"""Snakefile
+
+A snakemake workflow (snakepot) for calling the TPOT binary classifier:
+
+    rule copy: Copy the input dataset to the working directory
+    rule clean: Drop columns and rows with empty data
+    rule split: Create train, test, and validation data subsets
+    rule tpot: Run the TPOTClassifier with training and test data
+    rule evaluate: Evaluate model on validation data
+    rule predict: Predict unlabelled values in the input dataset 
+    rule end: Copy config.json to working directory to complete snakepot
+"""
+
+def expand_config(inlist):
+    """Quote argument lists to avoid errors in shell command.
+    Args:
+        inlist: A list of values to be converted to strings and quoted
+    Returns:
+        A list of quoted strings
+    """
+    return [ f"'{str(i)}'" for i in inlist ]
+
+# Set config file
+configfile: "config.json"
+# Create and set working directory
+workdir: config['directory']
+# Set snakefile directory
+snakedir: "../{}".format(config['directory'])
+
+# Set build target
+rule all:
+    input: "config.json"
+
+# STEP 0: Copy the input dataset into the working directory
+rule copy:
+    input: f"../{config['input']}"
+    output: "data.csv"
+    shell: "cp {input} data.csv"
+
+# STEP 1: Clean the dataset.
+rule clean:
+    # Read the dataset
+    input: "data.csv"
+    output: "cleaned.csv"
+    # Set parameters for columns to drop and one-hot encode
+    params:
+        drop=expand_config(config['drop_columns']),
+        encode=expand_config(config['encode_columns']),
+    # Call clean.py
+    shell:
+        "python ../src/clean.py --input {input} --drop {params.drop} --encode {params.encode}"
+
+rule split:
+    input: "cleaned.csv"
+    output: "training.csv", "test.csv", "unlabelled.csv"
+    params:
+        target_column=config['target_column'],
+        target_1=config['target_1'],
+        target_0=config['target_0'],
+        to_predict=config['to_predict'],
+        perc_split=config['perc_split'],
+    shell:
+        "python ../src/train_val_pred.py --cleaned {input} --target_column {params.target_column}"
+        " --target_1 {params.target_1} --target_0 {params.target_0} --to_predict {params.to_predict}"
+        " --perc_split {params.perc_split}"
+
+rule tpot:
+    input: "training.csv"
+    output: "tpot_pipe.py", "model.joblib"
+    params:
+        target=config['target_column'],
+        max=config['TPOT_max_time'],
+        outdir=config['directory'] # Note this argument is relative to /src
+    shell:
+        "python ../src/tpot_caller.py --training {input} --target {params.target} "
+        "--max_time {params.max} --outdir ../{params.outdir}"
+
+
+rule evaluate:
+    input:
+        test="test.csv",
+        training="training.csv"
+    output: "metrics.csv", "roc_data.csv", "precrec_data.csv"
+    params:
+        target=config['target_column'],
+    shell:
+        "python ../src/evaluate.py --test {input.test} --target {params.target} --training {input.training}"
+
+rule predict:
+    input: 
+        unlabelled="unlabelled.csv",
+        model="model.joblib",
+        training="training.csv"
+    output: "unlabelled_predictions.csv"
+    params: target=config['target_column']
+    shell:
+        "python ../src/predict.py --unlabelled {input.unlabelled} --training {input.training} --target {params.target}"
+
+rule end:
+    input: "unlabelled_predictions.csv"
+    output: "config.json"
+    shell:
+        "cp ../config.json ."

config.json

@@ -0,0 +1,17 @@
+{
+    "directory": "snakepot_test",
+    "input": "test/test.csv",
+    "drop_columns": [
+        "OO1", "OO2", "OO12", "OO12", "OO13", "OO14", "OO25", "OO28", "OO29", "OO35", "OO36",
+        "OO38", "OO37", "OO40", "OO39" 
+    ],
+    "encode_columns": [
+        "0045", "OO43", "OO44", "OO41", "OO42"
+    ],
+    "target_column": "target",
+    "target_1": "BENIGN",
+    "target_0": "DELETERIOUS",
+    "to_predict": "ND",
+    "perc_split": 0.2,
+    "TPOT_max_time": 5
+}