XAIMPL SUITE

Our solution uses the test results and ultimate diagnosis of hundreds of breast cancer patients and trains a model to predict whether a new sample is the case of a benign tumor or a malignant one. The aim is to have doctors not only rely on memory or even knowledge in order to correctly diagnose a patient. This model, combined with a doctor’s knowledge and experience can ultimately lead to diagnoses that are close to 100% correct.

The closing price of any stock can be predicted using this Long Short Term Memory Model. It has the ability to store some information in its "memory" and hence, utilize that in order to make predictions. The model is also explained using LIME, so that each prediction has an adequate explanation.

A Convolutional Neural Network is trained on a dataset consisting of thousands of images of moles or brown spots on the skin, in order to classify whether each is just a benign growth or a malignant melanoma. This type of application can be used as at a doctor's office or even at home.

Our unsupervised, machine learning solution takes 'The Movies Dataset' as input and uses movies' features like genre, keywords and actors to recommend movies that a similar to a given movie. Cosine Similarity is used to find the movies similar to each other.

Our solution uses the information about the movies watched and ratings given by the users to find similar users. Similar users are the ones which give same ratings to similar movies. The objective of collaborative filtering system is to recommend those movies to a user that are liked by users similar to him/her. This is an unsupervised, deep learning solution which is an implementation of autoencoders in Tensorflow.

Fake news is a major concern in our society right now. It has gone hand-in-hand with the rise of the data-driven era – not a coincidence when you consider the sheer volume of data we are generating every second! So in order to find out if a news is fake or real, we have created a fake news detector using LSTM model.

There could be thousands of factors which come into play when deciding whether a transaction is fraudulent. We use Genetic Algorithm to optimize the factors that are taken into account when classifying any credit card transaction as fraudulent

The timetable is optimized for staff-planning so the business can run smoothly, making sure the staff planning is as efficient as possible.

Our solution assigns various tags to entities in any text-based content. The state-of-the-art system uses deep contextualized embedded language modelings to cater to different contexts in which a specific entity may have been used.

Handwritten digit Recognition is becoming an important need as many services such as banking, postal address services etc are digitizing themselves in order to minimize human error. Our Model was trained on the MNIST data set using CNN to accurately predict handwritten digits.

A lot of data entry today is through handwritten documents such as prescriptions, invoices, tax documents, or forms. Our Model was trained on the Handwritten Digit model using transfer learning to incorporate characters along with digits.

With the help of this project, we can translate a word document in to any other language,to form a new document and we can also use the speech feature,for the document to be narrated.

Marketers spend a lot of time talking about the importance of getting the right behavior to the right people at the right time. To overcome this problem, our solution identifies who among your audience is most likely to make a purchase, accept an offer, sign up for a service, or to leave.

Understanding the sentiments of your customers or your competitor's customers plays a vital role in your business strategy.

The LSTM architecture is a classic solution to this problem. The Model is able to perform binary classification on text labeling as having positive or negative sentiment.

Applying sentiment analysis on reviews and comments/posts by people about a product over the internet can help gain insights about the current standing of a product among the masses.

BERT is a modern transformer well suited for high performance on a plethora of NLP tasks. The Model can label text into 3 or 5 classes representing the intensity of the sentiment from negative, neutral, positive.

Health care physicians often claim their service, medical fees amount from insurance organization, which they haven't provided. AI uses anomaly detection to offer fraud detection to insurance providers. This model will not only detect fraud claims but also the analytical parameters which play huge role in identification of pattern

Credit card frauds are one of the most common cyber security issues. Card not present transactions can lead to huge amount of transaction that are done by fraudulent person. Deep learning for catching the irrelevant pattern is used here. This model is used to detect fraudulent transactions and also genuine transaction

With the help of this solution, one can translate either English or French text to the other language. And for developing this solution, Encoder-Decoder RNN models are trained on packed and padded sequences of english-french sentences along with attention and masking. Moreover, torchtext is used as a NLP helper.

Diabetes is a common disease which often gives birth to myriad other medical conditions - which is why monitoring your blood glucose levels together with a number of other pertinent health factors on a regular basis is advisable. Our deep learning model uses your medical history to predict whether you are likely to contract diabetes in the coming 50 years or so.

According to dogtime.com, there are 266 different dog breeds. This number is also just one figure, other sources give their figures much higher than this and it is pretty tough to distinguish between these breeds. For example, its very difficult to identify a Whippet from an Italian Greyhound for a person. This model tries to predict different dog breeds on the go.

Are you annoyed of people scamming with similar documents again and again? This tool will help you finding such scams by calculating similarity indices of given documents.

According to BarkPost.com, there are more than 14 different breeds of Dogs and Cats which look the same. So, this model aims to classify them according to their respective species.

Human Action Recognition Using CNN models in real time videos to monitor human actions which could help in surveillance to identify suspicious activities and monitor activities in general to provide useful feedback

OpenAIMP's own chat bot on whats app, can answers queries about openAIMP organization.You can talk to the bot for general queries, in order to get quick response.