Accessories Recommendation Engine for NutriU Mobile APP

Pandas: For data manipulation, cleaning, and handling large datasets. Example: Merging customer data (device type, location, browsing history) into one comprehensive dataset.

NumPy: For numerical operations and handling arrays efficiently. Example: Matrix operations for recommendations.

Scikit-learn: For building machine learning models, especially for recommendation algorithms and quality control of the code. Example:KNeighborsClassifier() for collaborative filtering-based recommendation.; train_test_split() for splitting data into training and testing sets.; GridSearchCV() for hyperparameter tuning.

Surprise: A specialized library for building recommender systems, offering algorithms like collaborative filtering, matrix factorization, and more. Example: SVD() (Singular Value Decomposition) for matrix factorization-based recommendations.

TensorFlow/Keras: For building deep learning-based recommendation systems, especially if you want to implement neural collaborative filtering. Example: TensorFlow for creating custom neural network models to predict user-item interactions.

LightFM: A hybrid recommendation system library that allows for combining collaborative and content-based filtering. Example: Using LightFM to combine user browsing history with content metadata (e.g., product types, categories).

SciPy: For advanced statistical operations and distance calculations. Example: scipy.spatial.distance.cosine() for cosine similarity in collaborative filtering.

SQLAlchemy: For managing database queries if you are pulling data directly from SQL-based data sources. Example: Efficiently querying customer profiles, location, and browsing history.

Matplotlib/Seaborn: For visualizing data trends and recommendation accuracy. Example: Plotting model performance and customer interaction metrics.

Flask/Django: For deployment and creating REST APIs to serve the recommendation models in real-time. Example: Using Flask to create an API for serving accessory recommendations in the app.

Explanation:

Big query: Google fully managed serverless data warehouse, that had all customer interaction data using the airfrier device app. (Manage data, query using GoogleSQL - an SQL dialect, cross-cloud analytics, in-memory analysis service used to analyze large and complex datasets), visualize data from BigQuery by importing it in Data Studio - data visualization tool from Google)

Qlik view - Business Intelligence solution similar to Tableau/ Microstrategy

One-hot encoding: Converting categorical data to numerical data (1. integer encoding or 2. one-hot encoding).

- in sklearn.preprocessing (This encoding is needed for feeding categorical data to many scikit-learn estimators, notably linear models and SVMs with the standard kernels.)

Anonymization or Pseudonymization (Pseudonymisation is the process of replacing identifying information with random codes, which can be linked back to the original person with extra information, whereas anonymisation is the irreversible process of rendering personal data non-personal, and not subject to the GDPR.)

Recommender system:

- Collaborative filtering is a technique that can filter out items that a user might like on the basis of reactions by similar users. (Algorithms Based on K-Nearest Neighbours (k-NN)

- Content-Based Filtering: This method uses only information about the description and attributes of the items users has previously consumed to model user's preferences. In other words, these algorithms try to recommend items that are similar to those that a user liked in the past (or is examining in the present). In particular, various candidate items are compared with items previously rated by the user and the best-matching items are recommended.