v415ht313 / settyl_task Goto Github PK

##Data Preprocessing

Load the container event dataset from a JSON file. Preprocess the dataset by cleaning the external status descriptions and encoding the internal status labels.

##Model Architecture

The model architecture consists of a neural network with multiple dense layers. The input layer accepts vectorized external status descriptions. Hidden layers perform feature transformations. .The output layer predicts the internal status labels using softmax activation.

##Training Procedure

Split the preprocessed dataset into training and testing sets. Vectorize text data using TensorFlow's TextVectorization layer. Standardize features using StandardScaler. Compile the model with appropriate optimizer and loss function. Train the model on the training set for multiple epochs with mini-batch gradient descent. Evaluate the model on the testing set using accuracy, precision, and recall metrics.

##API Implementation

Develop an API using FastAPI framework. Define a request body model to accept external status descriptions as input. Implement an endpoint to make predictions based on the provided external status description. Load the trained model and necessary preprocessing components within the API.

##Testing Results

The trained model achieved a certain level of accuracy, precision, and recall on the testing set. The API endpoint successfully makes predictions based on the provided external status descriptions.