Product Group API

---

Clone repository

git clone https://github.com/AndranikUgujyan/product_group_classification.git

Install and create virtual environment

pip install virtualenv

virtualenv venv

Activate a virtual environment

source venv/bin/activate

Install required packages

pip3 install -r requirements.txt

---

Data:

We have balanced datasets.

To preprocess (normalize text) and split dataset into train/test run:

python -m product_model.data_proc.processor

The command will generate normal sampled datasets and will save as a csv files inside data folder.

python -m product_model.data_proc.embedding_processor

The command will generate embedded sampled datasets and will save as a csv files inside data folder.

Embedding This is a sentence-transformers model: It maps sentences & paragraphs to a 512 dimensional dense vector space and can be used for tasks like clustering or semantic search.

---

Models:

• Model 1: Multinomial Naive Bayes
• Model 2: Logistic Regression
• Model 3: Random Forest Classifier
• Model 3: GRU model
• Model 4: K Neighbors Classifier
• Model 5: 1D Convolutional Neural Network
• Model 6: Random Forest Classifier (on embedded data)
• Model 7: SVC (on embedded data)
• Model 8: Logistic Regression (on embedded data)
• Model 9: K Neighbors Classifier (on embedded data)
• Model 10: Decision Tree Classifier (on embedded data)
• Model 10: Gradient Boosting Classifier (on embedded data)

---

Model Train and Save

For train and save model based on normalized data run:

python -m product.training.train

For train and save model based on embedded data run:

 python -m product.training.train_embedded

---

Comparing the performance of each of models

Model's performances

Sorted model results by f1-score

---

Run API

---

We have 2 option to run API:

1. without docker
2. inside docker

---

Run API without docker

 python3 -m product_model.app

Test API without docker

curl  -X POST -d '{"data":[{"main_text": "USB 3.0 OTG STICK 64GB","add_text":"ULTRA DUAL 18__USB STICKS", "manufacturer":"SANDISK","model":"kn" }]'  http://172.17.0.2:8080/identify_product -H "Content-Type:application/json"

---

Run API inside docker docker

Build docker image

docker build -t product-classification-api:1.0 -f Dockerfile .

Run docker

docker run -e PORT=8080 --name imbd -p "7894:8080" product-classification-api:1.0

Test API in docker

curl  -X POST -d '{"data":[{"main_text": "USB 3.0 OTG STICK 64GB","add_text":"ULTRA DUAL 18__USB STICKS", "manufacturer":"SANDISK","model":"kn" }]'  http://127.0.0.1:7894/identify_product -H "Content-Type:application/json"

---

Here is an example response of identify_product prediction endpoint:

[
 "USB MEMORY"
]