TimeSeries Forecasting with ML Algorithms and there comparisons
summary
Compares concrete algorithms on a Covid19 kaggle univariate confirmed‑cases set using RMSE and shows ARIMA giving the lowest error RMSE 4000 while moving average and linear regression sit near 42,000 and 35,000, SVM around 22,000, LSTM ~10,000 and profit about 99,700. Implements naive, moving average, linear regression, SVM, RNN/LSTM, arima and profit, demonstrates train/validation by time split, and extends to multivariate VAR examples.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 0:47Exponential growth in pandemics
- 3:24Pandemic history and curve flattening
- 5:36Prediction accuracy metrics
- 6:42Prediction vs forecasting
- 7:51Naive and moving average algorithms
- 9:18Time-based train/validation splits
- 10:00Linear regression for forecasting
- 12:06Support Vector Machine for forecasting
- 15:17Recurrent Neural Networks
- 17:01RNN limitations and vanishing gradient
- 18:46LSTM networks and gates
- 20:36LSTM implementation and evaluation
- 20:59ARIMA and Prophet models
- 22:45Model results comparison
- 23:28Multivariate time series models
related talks
27mTime Series Forecasting on CoronaVirus Dataset
29mMachine Learning: Better than Traditional Techniques?
01h 26mApplying Statistical Modeling & Machine Learning to Perform Time-Series Forecasting
39mForecasting With Classical and Machine Learning Methods
29mTime Series Forecasting Techniques in Python
24mMachine learning and financial time series analysis
43mFeature Engineering for Time Series Forecasting
34mTime series for scikit-learn people