Skip to content

Latest commit

 

History

History
92 lines (65 loc) · 2.72 KB

Intel_Optimization.md

File metadata and controls

92 lines (65 loc) · 2.72 KB
Raw

Optimization with Intel oneAPI AI Analytics Toolkit

📝 Overview

This repository's stock price prediction model has been optimized using Intel's oneAPI AI Analytics Toolkit, specifically leveraging:

  • Intel Distribution for scikit-learn
  • Modin Distribution for parallelized Pandas operations
  • Intel oneDAAL library for accelerated machine learning algorithms

🚀 Key Optimizations

Intel scikit-learn Distribution

Utilized optimized algorithms for:

  • Linear Regression
  • Decision Trees
  • Random Forest

Intel scikit-learn vs Normal scikit-learn

Modin Distribution

Parallelized Pandas operations for:

  • Data loading and preprocessing
  • Data transformation and feature engineering

Pandas vs Modin

Intel oneDAAL Library

Accelerated machine learning algorithms for:

  • Principal Component Analysis (PCA)
  • K-Means Clustering
  • Linear Regression

🎯 Benefits

  • Improved Performance: Up to [30 and more]% reduction in training/inference time.
  • Enhanced Scalability: Efficiently handle large datasets and complex models.
  • Increased Accuracy: Optimized algorithms for improved prediction accuracy.

📋 Requirements

  • Intel oneAPI AI Analytics Toolkit installed
  • Compatible Intel hardware (e.g., Intel Core processors, Intel Xeon Scalable processors)

🛠️ Usage

  1. Clone the repository.
  2. Install Intel oneAPI AI Analytics Toolkit.
  3. Install Intel Distribution for scikit-learn and Modin.
  4. Build and run the optimized model using the provided instructions.
  5. Alternatively, you can install individual components using pip: 
    pip install scikit-learn-intelex
pip install modin[all]

💻 Code Snippets

# Import necessary libraries
from sklearnex import patch_sklearn
import modin.pandas as pd
from daal4py import PCA

# Patch scikit-learn to use Intel optimizations
patch_sklearn()

# Example: Load data using Modin
df = pd.read_csv('stock_prices.csv')

# Example: Preprocess data
df['Date'] = pd.to_datetime(df['Date'])
df.set_index('Date', inplace=True)

# Example: Train a Linear Regression model using Intel optimized scikit-learn
from sklearn.linear_model import LinearRegression

X = df[['Open', 'High', 'Low', 'Volume']]
y = df['Close']

model = LinearRegression()
model.fit(X, y)

# Example: Perform PCA using Intel oneDAAL
pca = PCA(n_components=2)
pca_result = pca.fit_transform(X)

print("PCA Result:", pca_result)

Python