lda模型:官方处理方式和自己处理数据对比

自己处理数据,然后分批训练,第一步先对比自己处理的方式和官方是否一致。

官方的代码

python 复制代码
import gensim
from gensim import corpora
from gensim.models import LdaModel

# 示例数据
documents = [
    "Human machine interface for lab abc computer applications",
    "A survey of user opinion of computer system response time",
    "The EPS user interface management system",
    "System and human system engineering testing of EPS",
    "Relation of user perceived response time to error measurement",
    "The generation of random binary unordered trees",
    "The intersection graph of paths in trees",
    "Graph minors IV Widths of trees and well quasi ordering",
    "Graph minors A survey"
]

# 预处理数据
texts = [[word for word in document.lower().split()] for document in documents]
dictionary = corpora.Dictionary(texts)
corpus = [dictionary.doc2bow(text) for text in texts]


# 训练 LDA 模型
lda_model = LdaModel(corpus, num_topics=5, id2word=dictionary, passes=15, random_state=2024)


# 打印每个主题的关键词
for idx, topic in lda_model.print_topics(-1):
    print(f"Topic: {idx}\nWords: {topic}\n")


# 推断新文档的主题分布
new_doc = "Human computer interaction"
new_doc_processed = [word for word in new_doc.lower().split()]
new_doc_bow = dictionary.doc2bow(new_doc_processed)
print(new_doc_bow)
print("New document topic distribution:", lda_model.get_document_topics(new_doc_bow))

结果

python 复制代码
Topic: 0
Words: 0.078*"graph" + 0.078*"trees" + 0.078*"the" + 0.078*"of" + 0.078*"in" + 0.078*"intersection" + 0.078*"paths" + 0.013*"minors" + 0.013*"interface" + 0.013*"survey"

Topic: 1
Words: 0.062*"of" + 0.034*"measurement" + 0.034*"relation" + 0.034*"to" + 0.034*"error" + 0.034*"perceived" + 0.034*"lab" + 0.034*"applications" + 0.034*"for" + 0.034*"machine"

Topic: 2
Words: 0.062*"minors" + 0.062*"trees" + 0.062*"the" + 0.062*"binary" + 0.062*"random" + 0.062*"generation" + 0.062*"unordered" + 0.062*"a" + 0.062*"survey" + 0.062*"graph"

Topic: 3
Words: 0.134*"system" + 0.073*"human" + 0.073*"eps" + 0.073*"and" + 0.073*"of" + 0.073*"engineering" + 0.073*"testing" + 0.012*"time" + 0.012*"user" + 0.012*"response"

Topic: 4
Words: 0.090*"of" + 0.090*"user" + 0.090*"system" + 0.049*"computer" + 0.049*"response" + 0.049*"time" + 0.049*"survey" + 0.049*"a" + 0.049*"interface" + 0.049*"management"

[(2, 1), (4, 1)]
New document topic distribution: [(0, 0.066698), (1, 0.7288686), (2, 0.06669144), (3, 0.06943816), (4, 0.068303764)]
python 复制代码
print(dictionary.token2id)

'''
{'abc': 0, 'applications': 1, 'computer': 2, 'for': 3, 'human': 4, 'interface': 5, 'lab': 6, 'machine': 7, 'a': 8, 'of': 9, 'opinion': 10, 'response': 11, 'survey': 12, 'system': 13, 'time': 14, 'user': 15, 'eps': 16, 'management': 17, 'the': 18, 'and': 19, 'engineering': 20, 'testing': 21, 'error': 22, 'measurement': 23, 'perceived': 24, 'relation': 25, 'to': 26, 'binary': 27, 'generation': 28, 'random': 29, 'trees': 30, 'unordered': 31, 'graph': 32, 'in': 33, 'intersection': 34, 'paths': 35, 'iv': 36, 'minors': 37, 'ordering': 38, 'quasi': 39, 'well': 40, 'widths': 41}
'''


print(corpus)

'''
[[(0, 1), (1, 1), (2, 1), (3, 1), (4, 1), (5, 1), (6, 1), (7, 1)], [(2, 1), (8, 1), (9, 2), (10, 1), (11, 1), (12, 1), (13, 1), (14, 1), (15, 1)], [(5, 1), (13, 1), (15, 1), (16, 1), (17, 1), (18, 1)], [(4, 1), (9, 1), (13, 2), (16, 1), (19, 1), (20, 1), (21, 1)], [(9, 1), (11, 1), (14, 1), (15, 1), (22, 1), (23, 1), (24, 1), (25, 1), (26, 1)], [(9, 1), (18, 1), (27, 1), (28, 1), (29, 1), (30, 1), (31, 1)], [(9, 1), (18, 1), (30, 1), (32, 1), (33, 1), (34, 1), (35, 1)], [(9, 1), (19, 1), (30, 1), (32, 1), (36, 1), (37, 1), (38, 1), (39, 1), (40, 1), (41, 1)], [(8, 1), (12, 1), (32, 1), (37, 1)]]
'''

自己处理方式

python 复制代码
def get_dictionary(input_data):
    output_dict = {}
    count = 0
    
    for l in input_data:
        l_list = l.strip().lower().split(" ")
        sorted_l_list = sorted(l_list)
        for k in sorted_l_list:
            if k not in output_dict:
                output_dict[k] = count
                count += 1
                
    return output_dict

my_dict = get_dictionary(documents)
print(my_dict)



def get_corpus(input_dict, input_data):
    output_list = []
    for l in input_data:
        tmp_dict = {}
        l_list = l.strip().lower().split(" ")
        for k in l_list:
            if k not in tmp_dict:
                tmp_dict[k] = 0
            tmp_dict[k] += 1
            
        tmp_list = []
        for k, v in tmp_dict.items():
            if k in input_dict.keys():
                tmp_list.append((input_dict[k], v))
            else:
                continue
        output_list.append(sorted(tmp_list))
        
    return output_list


my_corpus = get_corpus(my_dict, documents)
print(my_corpus)


def get_predict_corpus(input_dict, input_data):

    tmp_dict = {}
    l_list = input_data.strip().lower().split(" ")
    for k in l_list:
        if k not in tmp_dict:
            tmp_dict[k] = 0
        tmp_dict[k] += 1

    tmp_list = []
    for k, v in tmp_dict.items():
        if k in input_dict.keys():
            tmp_list.append((input_dict[k], v))
        else:
            continue

    return sorted(tmp_list)


'''
{'abc': 0, 'applications': 1, 'computer': 2, 'for': 3, 'human': 4, 'interface': 5, 'lab': 6, 'machine': 7, 'a': 8, 'of': 9, 'opinion': 10, 'response': 11, 'survey': 12, 'system': 13, 'time': 14, 'user': 15, 'eps': 16, 'management': 17, 'the': 18, 'and': 19, 'engineering': 20, 'testing': 21, 'error': 22, 'measurement': 23, 'perceived': 24, 'relation': 25, 'to': 26, 'binary': 27, 'generation': 28, 'random': 29, 'trees': 30, 'unordered': 31, 'graph': 32, 'in': 33, 'intersection': 34, 'paths': 35, 'iv': 36, 'minors': 37, 'ordering': 38, 'quasi': 39, 'well': 40, 'widths': 41}
[[(0, 1), (1, 1), (2, 1), (3, 1), (4, 1), (5, 1), (6, 1), (7, 1)], [(2, 1), (8, 1), (9, 2), (10, 1), (11, 1), (12, 1), (13, 1), (14, 1), (15, 1)], [(5, 1), (13, 1), (15, 1), (16, 1), (17, 1), (18, 1)], [(4, 1), (9, 1), (13, 2), (16, 1), (19, 1), (20, 1), (21, 1)], [(9, 1), (11, 1), (14, 1), (15, 1), (22, 1), (23, 1), (24, 1), (25, 1), (26, 1)], [(9, 1), (18, 1), (27, 1), (28, 1), (29, 1), (30, 1), (31, 1)], [(9, 1), (18, 1), (30, 1), (32, 1), (33, 1), (34, 1), (35, 1)], [(9, 1), (19, 1), (30, 1), (32, 1), (36, 1), (37, 1), (38, 1), (39, 1), (40, 1), (41, 1)], [(8, 1), (12, 1), (32, 1), (37, 1)]]
'''
python 复制代码
my_dict == dictionary.token2id

'''
True
'''

my_corpus == corpus

'''
True
'''
python 复制代码
# 训练 LDA 模型
my_lda_model = LdaModel(my_corpus, num_topics=5, passes=15, random_state=2024)
print(my_lda_model)


# 打印每个主题的关键词
for idx, topic in my_lda_model.print_topics(-1):
    print(f"Topic: {idx}\nWords: {topic}\n")


# 推断新文档的主题分布
new_doc = "Human computer interaction"
new_doc_bow = get_predict_corpus(my_dict, new_doc)
print(new_doc_bow)
print("New document topic distribution:", lda_model.get_document_topics(new_doc_bow))

结果

python 复制代码
LdaModel<num_terms=42, num_topics=5, decay=0.5, chunksize=2000>
Topic: 0
Words: 0.078*"32" + 0.078*"30" + 0.078*"18" + 0.078*"9" + 0.078*"33" + 0.078*"34" + 0.078*"35" + 0.013*"37" + 0.013*"5" + 0.013*"12"

Topic: 1
Words: 0.062*"9" + 0.034*"23" + 0.034*"25" + 0.034*"26" + 0.034*"22" + 0.034*"24" + 0.034*"6" + 0.034*"1" + 0.034*"3" + 0.034*"7"

Topic: 2
Words: 0.062*"37" + 0.062*"30" + 0.062*"18" + 0.062*"27" + 0.062*"29" + 0.062*"28" + 0.062*"31" + 0.062*"8" + 0.062*"12" + 0.062*"32"

Topic: 3
Words: 0.134*"13" + 0.073*"4" + 0.073*"16" + 0.073*"19" + 0.073*"9" + 0.073*"20" + 0.073*"21" + 0.012*"14" + 0.012*"15" + 0.012*"11"

Topic: 4
Words: 0.090*"9" + 0.090*"15" + 0.090*"13" + 0.049*"2" + 0.049*"11" + 0.049*"14" + 0.049*"12" + 0.049*"8" + 0.049*"5" + 0.049*"17"

[(2, 1), (4, 1)]
New document topic distribution: [(0, 0.06669798), (1, 0.72894156), (2, 0.06669143), (3, 0.06936743), (4, 0.06830162)]
相关推荐
测试者家园10 分钟前
基于DeepSeek和crewAI构建测试用例脚本生成器
人工智能·python·测试用例·智能体·智能化测试·crewai
大模型真好玩15 分钟前
准确率飙升!Graph RAG如何利用知识图谱提升RAG答案质量(四)——微软GraphRAG代码实战
人工智能·python·mcp
前端付豪22 分钟前
11、打造自己的 CLI 工具:从命令行到桌面效率神器
后端·python
前端付豪22 分钟前
12、用类写出更可控、更易扩展的爬虫框架🕷
后端·python
江太翁33 分钟前
Pytorch torch
人工智能·pytorch·python
网安INF1 小时前
深度学习中的逻辑回归:从原理到Python实现
人工智能·python·深度学习·算法·逻辑回归
青苔猿猿1 小时前
(5)aconda之.condarc文件配置
python·镜像源·.condarc文件
ningdisheng1 小时前
Python用图片生成banner.txt文件
python
0wioiw01 小时前
Python基础(吃洋葱小游戏)
开发语言·python·pygame
蓝婷儿1 小时前
Python 数据分析与可视化 Day 14 - 建模复盘 + 多模型评估对比(逻辑回归 vs 决策树)
python·数据分析·逻辑回归