Forecasting residual dye concentration in dye adsorption by activated carbon with Random Forest and Lag Features

Project info

• Field of studyChemistry and Materials
• LevelDoctoral dissertation (S3)
• Method typeTime Series
• Estimated build time± 7 working days
• Toolsscikit-learn RandomForestRegressor, pandas.DataFrame.shift
• PriceUS$31 - US$194

Research goal

Forecast residual dye concentration in dye adsorption by activated carbon with a lag-feature Random Forest and compare it with a simple baseline.

Variables

Example result

• 8.481RMSE recursive
• 8.817MAPE % recursive
• 7.648RMSE one-step
• -0.016R2 one-step
• 12Top lag (months)
• 0.56Top lag importance

The lag-feature random forest model forecasts residual dye concentration 12 periods ahead with RMSE 8.481 and MAPE 8.82%. The most important lag is t-12.

Computed from illustrative synthetic data, not from a real study.

About Random Forest with Lag Features

Random forest with lag features uses past values of the series as predictors, turning forecasting into tabular regression. It suits series with nonlinear patterns or many extra explanatory variables.

Steps

1. 1Create lag features (for example t-1, t-2, t-12) and calendar features
2. 2Split chronologically without shuffling
3. 3Train RandomForestRegressor
4. 4Evaluate MAE and RMSE on the test period
5. 5Compare with a naive forecast
6. 6Inspect lag importance and plot the result

Assumptions to check

• Splits must be chronological to avoid future leakage
• The pattern still holds in the forecast period
• Random forest does not extrapolate trends beyond the training range

Reading and reporting the results

Report MAE and RMSE against a naive baseline and the most important lags. Explain that this is one-step-ahead evaluation using actual lags.

Example Python code

A short starting script with this project's variable names. The delivered version is adapted to your data and fully commented.

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.ensemble import RandomForestRegressor

rng = np.random.default_rng(14)
n = 240
solution_ph = np.arange(n)
residual_dye_concentration = 100 + 0.2 * solution_ph + 12 * np.sin(2 * np.pi * solution_ph / 12) + rng.normal(0, 2, n)

df = pd.DataFrame({"y": residual_dye_concentration})
for k in (1, 2, 3, 12):
    df[f"lag{k}"] = df["y"].shift(k)
df = df.dropna().reset_index(drop=True)

split = len(df) - 24
X = df.drop(columns="y").values
y = df["y"].values
rf = RandomForestRegressor(n_estimators=300, random_state=14)
rf.fit(X[:split], y[:split])
pred = rf.predict(X[split:])
naive = df["lag1"].values[split:]

print("RF    MAE:", round(np.abs(y[split:] - pred).mean(), 3),
      "RMSE:", round(np.sqrt(((y[split:] - pred) ** 2).mean()), 3))
print("Naive MAE:", round(np.abs(y[split:] - naive).mean(), 3))
print(dict(zip(df.columns[1:], rf.feature_importances_.round(3))))

plt.plot(y[split:], label="Actual")
plt.plot(pred, label="RF")
plt.plot(naive, "--", label="Naive")
plt.legend()
plt.show()

Packages and prices

Choose how much help you need. Each higher package includes everything in the lower ones. Prices exclude 11% VAT.

• Payment: 50% down payment before work starts, balance after the deliverables are handed over
• Tax: Prices exclude 11% VAT
• Data and academic integrity: Results are computed from your own research data; you write and stand behind your manuscript, we help with the analysis and your understanding of it
• Revisions: Up to 2 revisions are included in the price (Package 3: 3 rounds)

Frequently asked questions