Python数据预测代写 | Project Proposal of Train Neural Network for Nature Languages


Project Aim
Implement a neural network based algorithm, which can classify some specific attribute via nature
languages and present the results with visualization techniques.
Project Objectives
• Collect and select appropriate dataset

• Preprocess dataset

• Train neural network via nature language

• Use neural network to classify predict attribute of books

• Visualization of data and result

Description of the project
The main aim of this project is to design a specific neural network to classify attribute(e.g genre)
via natural language(e.g. the summary of books). A neural network is a network of neurons
composed of artificial nodes. Normally, a neural network has at least three layers, input layer,
several hidden layers and output layer. This artificial network may used for classifying or predict
where it can be trained by dataset. This self-learning network can extract internal connections
from seemingly unrelated data and generate reasonable predictions through extensive training.
Thus it is theoretically feasible to use neural network to classify the some attribute of books via
nature languages.

In today’s world, Amazon and other bookstores need to make a preliminary of the books which
will be sold in their store, in order to maintain their good reviews. Usually they invite professional
personnel to classify books after reading, but this method is very inefficient when faced with a
large number of books. So they need an algorithm that can handle a large number of book rating
to assist in their work. In many news sites, they also use natural language classify technics to
calculate the category of the corresponding news, which could appear accurately to users on the
pages they browse.

The main point in this project is use efficient and reasonable algorithm to train classify model. This
model need to be trained by nature language and numbers. Now I assume that I need to spend
most of the time in program nature language identify algorithms and select the best from them.

Initial area of research
In the first stage, preprocessing dataset is essential, most of machine learning algorithms require
data to be input in the form of matrixes. When I writing this proposal, I think that classify neural
networks(RNN or LSTM) are more incline to supervised learning, so the original dataset needs to
be divided into training set and test set. This step can greatly improved the efficiency and

versatility of algorithm sorting with the help of programming language. At the last step of
preprocessing dataset, presenting the abstract and discrete matrix in the form of charts, this step
will be very helpful for future work.

Expected outcomes
Outcomes of this project is a mature neural network which has been train by thousands times.
This algorithm could used to classify the specific attribute via natural language. All results from
this algorithm are clearly presented in a visualization method. The accuracy of the algorithm will
be maximized during programming process(it is currently expected to be at least 70%).