Neural Network Model Training in Keras
Budget: $1,500 – $3,000 USD
Objective:
We are seeking an individual or company skilled in neural networks and Keras to develop a predictive model. This model will forecast the occurrence of a phenomenon based on historical data.
Requirements:
Data Acquisition: Historical data required for training the model will be fetched via an API from our comprehensive database, which is updated multiple times weekly.
Data Processing: All retrieved data will be processed to facilitate effective model training. This includes Exploratory Data Analysis (EDA) to identify impactful data, outlier detection for data quality, and feature engineering to determine the final model inputs.
Model Development: Train two types of models using Deep Neural Network (DNN) and Long Short-Term Memory (LSTM) architectures:
A classification model to determine the likelihood of an event's occurrence.
A prediction model to quantify the value of the event.
Implementation Platform: The system will be implemented on a cloud platform, such as Digital Ocean or Google Cloud Platform.
Web Interface: A user-friendly web interface for system management will be provided, enabling model management, testing, and real-time accuracy assessments.
Operational Details:
Data Integration: Seamless integration with our database API to continuously retrieve and update past data.
Data Handling Module:
Conduct Exploratory Data Analysis (EDA).
Detect and eliminate data outliers.
Perform final feature engineering.
Model Training Module:
Aim to achieve approximately 85% accuracy in models.
Train models under two architectures: DNN and LSTM for both classification and quantitative prediction tasks.
Cloud Integration:
Migrate the entire solution to a cloud platform.
Provide APIs for ongoing training and prediction.
Implement an automatic scheduling mechanism for model updates when new data is added to the database.
Web Management Interface:
Facilitate remote model management, including swapping active models and initiating predictions.
Provide tools for login and integration with cloud IAM and cloud functions.
Enable model management through cloud storage, with the model name sent as an API parameter.
Host an external database for statistical comparison of predictions against actual outcomes and display statistical charts.
We are seeking an individual or company skilled in neural networks and Keras to develop a predictive model. This model will forecast the occurrence of a phenomenon based on historical data.
Requirements:
Data Acquisition: Historical data required for training the model will be fetched via an API from our comprehensive database, which is updated multiple times weekly.
Data Processing: All retrieved data will be processed to facilitate effective model training. This includes Exploratory Data Analysis (EDA) to identify impactful data, outlier detection for data quality, and feature engineering to determine the final model inputs.
Model Development: Train two types of models using Deep Neural Network (DNN) and Long Short-Term Memory (LSTM) architectures:
A classification model to determine the likelihood of an event's occurrence.
A prediction model to quantify the value of the event.
Implementation Platform: The system will be implemented on a cloud platform, such as Digital Ocean or Google Cloud Platform.
Web Interface: A user-friendly web interface for system management will be provided, enabling model management, testing, and real-time accuracy assessments.
Operational Details:
Data Integration: Seamless integration with our database API to continuously retrieve and update past data.
Data Handling Module:
Conduct Exploratory Data Analysis (EDA).
Detect and eliminate data outliers.
Perform final feature engineering.
Model Training Module:
Aim to achieve approximately 85% accuracy in models.
Train models under two architectures: DNN and LSTM for both classification and quantitative prediction tasks.
Cloud Integration:
Migrate the entire solution to a cloud platform.
Provide APIs for ongoing training and prediction.
Implement an automatic scheduling mechanism for model updates when new data is added to the database.
Web Management Interface:
Facilitate remote model management, including swapping active models and initiating predictions.
Provide tools for login and integration with cloud IAM and cloud functions.
Enable model management through cloud storage, with the model name sent as an API parameter.
Host an external database for statistical comparison of predictions against actual outcomes and display statistical charts.